From eafb3938a21a75dc45fc9e191672472f42ed051a Mon Sep 17 00:00:00 2001 From: GStechschulte Date: Wed, 1 Nov 2023 20:38:10 +0100 Subject: [PATCH] replace bambi.plots with bmb.interpret. --- docs/notebooks/plot_comparisons.ipynb | 3708 +++++++++++++------------ 1 file changed, 1868 insertions(+), 1840 deletions(-) diff --git a/docs/notebooks/plot_comparisons.ipynb b/docs/notebooks/plot_comparisons.ipynb index 7df49b7e7..074075051 100644 --- a/docs/notebooks/plot_comparisons.ipynb +++ b/docs/notebooks/plot_comparisons.ipynb @@ -1,1897 +1,1925 @@ { - "cells": [ - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "# Plot Comparisons\n", - "\n", - "`comparisons` and `plot_comparisons` are a part of Bambi's sub-package `plots` that feature a set of functions used to interpret complex regression models. This sub-package is inspired by the R package [marginaleffects](https://vincentarelbundock.github.io/marginaleffects/articles/predictions.html#conditional-adjusted-predictions-plot). These two functions allow the modeler to **compare** the predictions made by a model for different contrast and covariate values. Below, it is described why comparing predictions is useful in interpreting generalized linear models (GLMs), how this methodology is implemented in Bambi, and how to use `comparisons` and `plot_comparisons`. It is assumed that the reader is familiar with the basics of GLMs. If not, refer to the Bambi [Basic Building Blocks](https://bambinos.github.io/bambi/notebooks/how_bambi_works.html#Link-functions) example.\n", - "\n", - "Due to the link function in a GLM, there are typically three quantities of interest to interpret:\n", - "\n", - "1. the linear predictor $\\eta = X\\beta$ where $X$ is an $n$ x $p$ matrix of explanatory variables.\n", - "2. the mean $\\mu = g^{-1}(\\eta)$ where the link function $g(\\cdot)$ relates the linear predictor to the mean of the outcome variable $\\mu = g^{-1}(\\eta) = g^{-1}(X\\beta)$\n", - "3. the response variable $Y \\sim \\mathcal{D}(\\mu, \\theta)$ where $\\mu$ is the mean parameter and $\\theta$ is (possibly) a vector that contains all the other \"auxillary\" parameters of the distribution.\n", - " \n", - "Often, with GLMs, $\\eta$ is linear in the parameters, but nonlinear in relation of inputs to the outcome $Y$ due to the link function $g$. Thus, as modelers, we are usually more interested in interpreting (2) and (3). For example, in logistic regression, the linear predictor is on the log-odds scale, but the quantity of interest is on the probability scale. In Poisson regression, the linear predictor is on the log-scale, but the response variable is on the count scale. Referring back to logistic regression, a specified difference in one of the $x$ variables does _not_ correspond to a constant difference in the the probability of the outcome.\n", - "\n", - "It is often helpful with GLMs, for the modeler and audience, to have a summary that gives the expected difference in the outcome corresponding to a unit difference in each of the input variables. Thus, the goal of `comparisons` and `plot_comparisons` is to provide the modeler with a summary and visualization of the average predicted difference." + "cells": [ + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "# Plot Comparisons\n", + "\n", + "`comparisons` and `plot_comparisons` are a part of Bambi's sub-package `plots` that feature a set of functions used to interpret complex regression models. This sub-package is inspired by the R package [marginaleffects](https://vincentarelbundock.github.io/marginaleffects/articles/predictions.html#conditional-adjusted-predictions-plot). These two functions allow the modeler to **compare** the predictions made by a model for different contrast and covariate values. Below, it is described why comparing predictions is useful in interpreting generalized linear models (GLMs), how this methodology is implemented in Bambi, and how to use `comparisons` and `plot_comparisons`. It is assumed that the reader is familiar with the basics of GLMs. If not, refer to the Bambi [Basic Building Blocks](https://bambinos.github.io/bambi/notebooks/how_bambi_works.html#Link-functions) example.\n", + "\n", + "Due to the link function in a GLM, there are typically three quantities of interest to interpret:\n", + "\n", + "1. the linear predictor $\\eta = X\\beta$ where $X$ is an $n$ x $p$ matrix of explanatory variables.\n", + "2. the mean $\\mu = g^{-1}(\\eta)$ where the link function $g(\\cdot)$ relates the linear predictor to the mean of the outcome variable $\\mu = g^{-1}(\\eta) = g^{-1}(X\\beta)$\n", + "3. the response variable $Y \\sim \\mathcal{D}(\\mu, \\theta)$ where $\\mu$ is the mean parameter and $\\theta$ is (possibly) a vector that contains all the other \"auxillary\" parameters of the distribution.\n", + " \n", + "Often, with GLMs, $\\eta$ is linear in the parameters, but nonlinear in relation of inputs to the outcome $Y$ due to the link function $g$. Thus, as modelers, we are usually more interested in interpreting (2) and (3). For example, in logistic regression, the linear predictor is on the log-odds scale, but the quantity of interest is on the probability scale. In Poisson regression, the linear predictor is on the log-scale, but the response variable is on the count scale. Referring back to logistic regression, a specified difference in one of the $x$ variables does _not_ correspond to a constant difference in the the probability of the outcome.\n", + "\n", + "It is often helpful with GLMs, for the modeler and audience, to have a summary that gives the expected difference in the outcome corresponding to a unit difference in each of the input variables. Thus, the goal of `comparisons` and `plot_comparisons` is to provide the modeler with a summary and visualization of the average predicted difference." + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Average Predictive Differences\n", + "\n", + "Here, we adopt the notation from Chapter 14.4 of [Regression and Other Stories](https://avehtari.github.io/ROS-Examples/) to describe average predictive differences. Assume we have fit a Bambi model predicting an outcome $Y$ based on inputs $X$ and parameters $\\theta$. Consider the following scalar inputs:\n", + "\n", + "$$w: \\text{the input of interest}$$\n", + "$$c: \\text{all the other inputs}$$\n", + "$$X = (w, c)$$\n", + "\n", + "Suppose for the input of interest, we are interested in comparing $w^{\\text{high}}$ to $w^{\\text{low}}$ (perhaps age = $60$ and $40$ respectively) with all other inputs $c$ held constant. The _predictive difference_ in the outcome changing **only** $w$ is:\n", + "\n", + "$$\\text{average predictive difference} = \\mathbb{E}(y|w^{\\text{high}}, c, \\theta) - \\mathbb{E}(y|w^{\\text{low}}, c, \\theta)$$\n", + "\n", + "Selecting the maximum and minimum values of $w$ and averaging over all other inputs $c$ in the data gives you a new \"hypothetical\" dataset and corresponds to counting all pairs of transitions of $(w^\\text{low})$ to $(w^\\text{high})$, i.e., differences in $w$ with $c$ held constant. The difference between these two terms is the average predictive difference." + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Computing Average Predictive Differences\n", + "\n", + "The objective of `comparisons` and `plot_comparisons` is to compute the expected difference in the outcome corresponding to three different scenarios for $w$ and $c$ where $w$ is either provided by the user, else a default value is computed by Bambi (described in the default values section). The three scenarios are:\n", + "\n", + "1. user provided values for $c$.\n", + "2. a grid of equally spaced and central values for $c$.\n", + "3. empirical distribution (original data used to fit the model) for $c$. \n", + "\n", + "In the case of (1) and (2) above, Bambi assembles all pairwise combinations (transitions) of $w$ and $c$ into a new \"hypothetical\" dataset. In (3), Bambi uses the original $c$, but replaces $w$ with the user provided value or the default value computed by Bambi. In each scenario, predictions are made on the data using the fitted model. Once the predictions are made, comparisons are computed using the posterior samples by taking the difference in the predicted outcome for each pair of transitions. The average of these differences is the average predictive difference. \n", + "\n", + "Thus, the goal of `comparisons` and `plot_comparisons` is to provide the modeler with a summary and visualization of the average predictive difference. Below, we demonstrate how to compute and plot average predictive differences with `comparisons` and `plot_comparions` using several examples." + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": {}, + "outputs": [], + "source": [ + "import arviz as az\n", + "import numpy as np\n", + "import pandas as pd\n", + "\n", + "import bambi as bmb" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Zero Inflated Poisson" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "We model and predict how many fish are caught by visitors at a state park using survey [data](\"https://stats.idre.ucla.edu/stat/data/fish.csv\"). Many visitors catch zero fish, either because they did not fish at all, or because they were unlucky. We would like to explicitly model this bimodal behavior (zero versus non-zero) using a Zero Inflated Poisson model, and to compare how different inputs of interest $w$ and other covariate values $c$ are associated with the number of fish caught. The dataset contains data on 250 groups that went to a state park to fish. Each group was questioned about how many fish they caught (`count`), how many children were in the group (`child`), how many people were in the group (`persons`), if they used a live bait and whether or not they brought a camper to the park (`camper`)." + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "metadata": {}, + "outputs": [], + "source": [ + "fish_data = pd.read_stata(\"http://www.stata-press.com/data/r11/fish.dta\")\n", + "cols = [\"count\", \"livebait\", \"camper\", \"persons\", \"child\"]\n", + "fish_data = fish_data[cols]\n", + "fish_data[\"livebait\"] = pd.Categorical(fish_data[\"livebait\"])\n", + "fish_data[\"camper\"] = pd.Categorical(fish_data[\"camper\"])" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Auto-assigning NUTS sampler...\n", + "Initializing NUTS using jitter+adapt_diag...\n", + "Multiprocess sampling (4 chains in 4 jobs)\n", + "NUTS: [count_psi, Intercept, livebait, camper, persons, child]\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + " |████████████████████████████████| 100.00% [8000/8000 00:03<00:00 Sampling 4 chains, 0 divergences]\r" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 4 seconds.\n" + ] + } + ], + "source": [ + "fish_model = bmb.Model(\n", + " \"count ~ livebait + camper + persons + child\", \n", + " fish_data, \n", + " family='zero_inflated_poisson'\n", + ")\n", + "\n", + "fish_idata = fish_model.fit(\n", + " draws=1000, \n", + " target_accept=0.95, \n", + " random_seed=1234, \n", + " chains=4\n", + ")" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### User Provided Values\n", + "\n", + "First, an example of scenario 1 (user provided values) is given below. In both `plot_comparisons` and `comparisons`, $w$ and $c$ are represented by `contrast` and `conditional`, respectively. The modeler has the ability to pass their own values for `contrast` and `conditional` by using a dictionary where the key-value pairs are the covariate and value(s) of interest. For example, if we wanted to compare the number of fish caught for $4$ versus $1$ `persons` conditional on a range of `child` and `livebait` values, we would pass the following dictionary in the code block below. By default, for $w$, Bambi compares $w^\\text{high}$ to $w^\\text{low}$. Thus, in this example, $w^\\text{high}$ = 4 and $w^\\text{low}$ = 1. The user is not limited to passing a list for the values. A `np.array` can also be used. Furthermore, Bambi by default, maps the order of the dict keys to the main, group, and panel of the matplotlib figure. Below, since `child` is the first key, this is used for the x-axis, and `livebait` is used for the group (color). If a third key was passed, it would be used for the panel (facet)." + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "## Average Predictive Differences\n", + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fig, ax = bmb.interpret.plot_comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast={\"persons\": [1, 4]},\n", + " conditional={\"child\": [0, 1, 2], \"livebait\": [0, 1]},\n", + ") \n", + "fig.set_size_inches(7, 3)" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "The plot above shows that, comparing $4$ to $1$ persons given $0$ children and using livebait, the expected difference is about $26$ fish. When not using livebait, the expected difference decreases substantially to about $5$ fish. Using livebait with a group of people is associated with a much larger expected difference in the number of fish caught. \n", + "\n", + "`comparisons` can be called to view a summary dataframe that includes the term $w$ and its contrast, the specified `conditional` covariate, and the expected difference in the outcome with the uncertainty interval (by default the 94% highest density interval is computed). " + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevaluechildlivebaitcamperestimatelower_3.0%upper_97.0%
0personsdiff(1.0, 4.0)0.00.01.04.8344722.5634727.037150
1personsdiff(1.0, 4.0)0.01.01.026.42318823.73972929.072748
2personsdiff(1.0, 4.0)1.00.01.01.2020030.6316291.780965
3personsdiff(1.0, 4.0)1.01.01.06.5719435.4692757.642248
4personsdiff(1.0, 4.0)2.00.01.00.3013840.1436760.467608
5personsdiff(1.0, 4.0)2.01.01.01.6484171.1404152.187190
\n", + "
" + ], + "text/plain": [ + " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", + "0 persons diff (1.0, 4.0) ... 4.834472 2.563472 7.037150\n", + "1 persons diff (1.0, 4.0) ... 26.423188 23.739729 29.072748\n", + "2 persons diff (1.0, 4.0) ... 1.202003 0.631629 1.780965\n", + "3 persons diff (1.0, 4.0) ... 6.571943 5.469275 7.642248\n", + "4 persons diff (1.0, 4.0) ... 0.301384 0.143676 0.467608\n", + "5 persons diff (1.0, 4.0) ... 1.648417 1.140415 2.187190\n", "\n", - "Selecting the maximum and minimum values of $w$ and averaging over all other inputs $c$ in the data gives you a new \"hypothetical\" dataset and corresponds to counting all pairs of transitions of $(w^\\text{low})$ to $(w^\\text{high})$, i.e., differences in $w$ with $c$ held constant. The difference between these two terms is the average predictive difference." + "[6 rows x 9 columns]" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### Computing Average Predictive Differences\n", + "execution_count": 5, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast={\"persons\": [1, 4]},\n", + " conditional={\"child\": [0, 1, 2], \"livebait\": [0, 1]},\n", + ") " + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "But why is `camper` also in the summary dataframe? This is because in order to peform predictions, Bambi is expecting a value for each covariate used to fit the model. Additionally, with GLM models, average predictive comparisons are conditional in the sense that the estimate depends on the values of all the covariates in the model. Thus, for unspecified covariates, `comparisons` and `plot_comparisons` computes a default value (mean or mode based on the data type of the covariate). Thus, $c$ = `child`, `livebait`, `camper`. Each row in the summary dataframe is read as \"comparing $4$ to $1$ persons conditional on $c$, the expected difference in the outcome is $y$.\"" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Multiple contrast values\n", + "\n", + "Users can also perform comparisons on multiple contrast values. For example, if we wanted to compare the number of fish caught between $(1, 2)$, $(1, 4)$, and $(2, 4)$ `persons` conditional on a range of values for `child` and `livebait`." + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevaluechildlivebaitcamperestimatelower_3.0%upper_97.0%
0personsdiff(1, 2)0.00.01.00.5276270.2954510.775465
1personsdiff(1, 2)0.01.01.02.8836942.6056903.177685
2personsdiff(1, 2)1.00.01.00.1313190.0673390.195132
3personsdiff(1, 2)1.01.01.00.7179650.5929680.857893
4personsdiff(1, 2)2.00.01.00.0329600.0152120.052075
5personsdiff(1, 2)2.01.01.00.1802700.1231730.244695
6personsdiff(1, 4)0.00.01.04.8344722.5634727.037150
7personsdiff(1, 4)0.01.01.026.42318823.73972929.072748
8personsdiff(1, 4)1.00.01.01.2020030.6316291.780965
9personsdiff(1, 4)1.01.01.06.5719435.4692757.642248
10personsdiff(1, 4)2.00.01.00.3013840.1436760.467608
11personsdiff(1, 4)2.01.01.01.6484171.1404152.187190
12personsdiff(2, 4)0.00.01.04.3068452.2670976.280005
13personsdiff(2, 4)0.01.01.023.53949420.99093126.240169
14personsdiff(2, 4)1.00.01.01.0706830.5659311.585718
15personsdiff(2, 4)1.01.01.05.8539784.8589576.848519
16personsdiff(2, 4)2.00.01.00.2684230.1240330.412274
17personsdiff(2, 4)2.01.01.01.4681471.0248001.960934
\n", + "
" + ], + "text/plain": [ + " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", + "0 persons diff (1, 2) ... 0.527627 0.295451 0.775465\n", + "1 persons diff (1, 2) ... 2.883694 2.605690 3.177685\n", + "2 persons diff (1, 2) ... 0.131319 0.067339 0.195132\n", + "3 persons diff (1, 2) ... 0.717965 0.592968 0.857893\n", + "4 persons diff (1, 2) ... 0.032960 0.015212 0.052075\n", + "5 persons diff (1, 2) ... 0.180270 0.123173 0.244695\n", + "6 persons diff (1, 4) ... 4.834472 2.563472 7.037150\n", + "7 persons diff (1, 4) ... 26.423188 23.739729 29.072748\n", + "8 persons diff (1, 4) ... 1.202003 0.631629 1.780965\n", + "9 persons diff (1, 4) ... 6.571943 5.469275 7.642248\n", + "10 persons diff (1, 4) ... 0.301384 0.143676 0.467608\n", + "11 persons diff (1, 4) ... 1.648417 1.140415 2.187190\n", + "12 persons diff (2, 4) ... 4.306845 2.267097 6.280005\n", + "13 persons diff (2, 4) ... 23.539494 20.990931 26.240169\n", + "14 persons diff (2, 4) ... 1.070683 0.565931 1.585718\n", + "15 persons diff (2, 4) ... 5.853978 4.858957 6.848519\n", + "16 persons diff (2, 4) ... 0.268423 0.124033 0.412274\n", + "17 persons diff (2, 4) ... 1.468147 1.024800 1.960934\n", "\n", - "Thus, the goal of `comparisons` and `plot_comparisons` is to provide the modeler with a summary and visualization of the average predictive difference. Below, we demonstrate how to compute and plot average predictive differences with `comparisons` and `plot_comparions` using several examples." + "[18 rows x 9 columns]" ] }, - { - "cell_type": "code", - "execution_count": 1, - "metadata": {}, - "outputs": [], - "source": [ - "import arviz as az\n", - "import numpy as np\n", - "import pandas as pd\n", + "execution_count": 6, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "multiple_values = bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast={\"persons\": [1, 2, 4]},\n", + " conditional={\"child\": [0, 1, 2], \"livebait\": [0, 1]}\n", + ")\n", + "\n", + "multiple_values" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Notice how the contrast $w$ varies while the covariates $c$ are held constant. Currently, however, plotting multiple contrast values can be difficult to interpret since the contrast is \"abstracted\" away onto the y-axis. Thus, it would be difficult to interpret which portion of the plot corresponds to which contrast value. Therefore, it is currently recommended that if you want to plot multiple contrast values, call `comparisons` directly to obtain the summary dataframe and plot the results yourself." + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Default contrast and conditional values\n", + "\n", + "Now, we move onto scenario 2 described above (grid of equally spaced and central values) in computing average predictive comparisons. You are not required to pass values for `contrast` and `conditional`. If you do not pass values, Bambi will compute default values for you. Below, it is described how these default values are computed.\n", + "\n", + "The default value for `contrast` is a _centered difference_ at the mean for a contrast variable with a numeric dtype, and _unique levels_ for a contrast varaible with a categorical dtype. For example, if the modeler is interested in the comparison of a $5$ unit increase in $w$ where $w$ is a numeric variable, Bambi computes the mean and then subtracts and adds $2.5$ units to the mean to obtain a _centered difference_. By default, if no value is passed for the contrast covariate, Bambi computes a one unit centered difference at the mean. For example, if only `contrast=\"persons\"` is passed, then $\\pm$ $0.5$ is applied to the mean of persons. If $w$ is a categorical variable, Bambi computes and returns the unique levels. For example, if $w$ has levels [\"high scool\", \"vocational\", \"university\"], Bambi computes and returns the unique values of this variable.\n", + "\n", + "The default values for `conditional` are more involved. Currently, by default, if a dict or list is passed to `conditional`, Bambi uses the ordering (keys if dict and elements if list) to determine which covariate to use as the main, group (color), and panel (facet) variable. This is the same logic used in `plot_comparisons` described above. Subsequently, the default values used for the `conditional` covariates depend on their ordering **and** dtype. Below, the psuedocode used for computing default values covariates passed to `conditional` is outlined:\n", + "\n", + "```python\n", + "if v == \"main\":\n", + " \n", + " if v == numeric:\n", + " return np.linspace(v.min(), v.max(), 50)\n", + " elif v == categorical:\n", + " return np.unique(v)\n", + "\n", + "elif v == \"group\":\n", + " \n", + " if v == numeric:\n", + " return np.quantile(v, np.linspace(0, 1, 5))\n", + " elif v == categorical:\n", + " return np.unique(v)\n", + "\n", + "elif v == \"panel\":\n", + " \n", + " if v == numeric:\n", + " return np.quantile(v, np.linspace(0, 1, 5))\n", + " elif v == categorical:\n", + " return np.unique(v)\n", + "```\n", + "\n", + "Thus, letting Bambi compute default values for `conditional` is equivalent to creating a hypothetical \"data grid\" of new values. Lets say we are interested in comparing the number of fish caught for the contrast `livebait` conditional on `persons` and `child`. This time, lets call `comparisons` first to gain an understanding of the data generating the plot." + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevaluepersonschildcamperestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)1.0000000.01.01.6946461.2528032.081207
1livebaitdiff(0.0, 1.0)1.0000001.01.00.4224480.2990520.551766
2livebaitdiff(0.0, 1.0)1.0000003.01.00.0269230.0127520.043035
3livebaitdiff(0.0, 1.0)1.0612240.01.01.7874121.3429792.203158
4livebaitdiff(0.0, 1.0)1.0612241.01.00.4455550.3172530.580117
5livebaitdiff(0.0, 1.0)1.0612243.01.00.0283930.0134520.045276
6livebaitdiff(0.0, 1.0)1.1224490.01.01.8852701.4229372.313218
7livebaitdiff(0.0, 1.0)1.1224491.01.00.4699290.3353730.609249
8livebaitdiff(0.0, 1.0)1.1224493.01.00.0299440.0141650.047593
9livebaitdiff(0.0, 1.0)1.1836740.01.01.9885001.5016502.424762
\n", + "
" ], - "source": [ - "fish_model = bmb.Model(\n", - " \"count ~ livebait + camper + persons + child\", \n", - " fish_data, \n", - " family='zero_inflated_poisson'\n", - ")\n", + "text/plain": [ + " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", + "0 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", + "1 livebait diff (0.0, 1.0) ... 0.422448 0.299052 0.551766\n", + "2 livebait diff (0.0, 1.0) ... 0.026923 0.012752 0.043035\n", + "3 livebait diff (0.0, 1.0) ... 1.787412 1.342979 2.203158\n", + "4 livebait diff (0.0, 1.0) ... 0.445555 0.317253 0.580117\n", + "5 livebait diff (0.0, 1.0) ... 0.028393 0.013452 0.045276\n", + "6 livebait diff (0.0, 1.0) ... 1.885270 1.422937 2.313218\n", + "7 livebait diff (0.0, 1.0) ... 0.469929 0.335373 0.609249\n", + "8 livebait diff (0.0, 1.0) ... 0.029944 0.014165 0.047593\n", + "9 livebait diff (0.0, 1.0) ... 1.988500 1.501650 2.424762\n", "\n", - "fish_idata = fish_model.fit(\n", - " draws=1000, \n", - " target_accept=0.95, \n", - " random_seed=1234, \n", - " chains=4\n", - ")" + "[10 rows x 9 columns]" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### User Provided Values\n", - "\n", - "First, an example of scenario 1 (user provided values) is given below. In both `plot_comparisons` and `comparisons`, $w$ and $c$ are represented by `contrast` and `conditional`, respectively. The modeler has the ability to pass their own values for `contrast` and `conditional` by using a dictionary where the key-value pairs are the covariate and value(s) of interest. For example, if we wanted to compare the number of fish caught for $4$ versus $1$ `persons` conditional on a range of `child` and `livebait` values, we would pass the following dictionary in the code block below. By default, for $w$, Bambi compares $w^\\text{high}$ to $w^\\text{low}$. Thus, in this example, $w^\\text{high}$ = 4 and $w^\\text{low}$ = 1. The user is not limited to passing a list for the values. A `np.array` can also be used. Furthermore, Bambi by default, maps the order of the dict keys to the main, group, and panel of the matplotlib figure. Below, since `child` is the first key, this is used for the x-axis, and `livebait` is used for the group (color). If a third key was passed, it would be used for the panel (facet)." + "execution_count": 7, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "contrast_df = bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=[\"persons\", \"child\"],\n", + ")\n", + "\n", + "contrast_df.head(10)" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "As `livebait` was encoded as a categorical dtype, Bambi returned the unique levels of $[0, 1]$ for the contrast. `persons` and `child` were passed as the first and second element and thus act as the main and group variables, respectively. It can be see from the output above, that an equally spaced grid was used to compute the values for `persons`, whereas a quantile based grid was used for `child`. Furthermore, as `camper` was unspecified, the mode was used as the default value. Lets go ahead and plot the commparisons." + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" ] }, - { - "cell_type": "code", - "execution_count": 4, - "metadata": {}, - "outputs": [ - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "fig, ax = plot_comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast={\"persons\": [1, 4]},\n", - " conditional={\"child\": [0, 1, 2], \"livebait\": [0, 1]},\n", - ") \n", - "fig.set_size_inches(7, 3)" + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fig, ax = bmb.interpret.plot_comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=[\"persons\", \"child\"],\n", + ") \n", + "fig.set_size_inches(7, 3)" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "The plot shows us that the expected differences in fish caught comparing a group of people who use livebait and no livebait is not only conditional on the number of persons, but also children. However, the plotted comparisons for `child` = $3$ is difficult to interpret on a single plot. Thus, it can be useful to pass specific `group` and `panel` arguments to aid in the interpretation of the plot. Therefore, `subplot_kwargs` allows the user to manipulate the plotting by passing a dictionary where the keys are `{\"main\": ..., \"group\": ..., \"panel\": ...}` and the values are the names of the covariates to be plotted. Below, we plot the same comparisons as above, but this time we specify `group` and `panel` to both be `child`." + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "The plot above shows that, comparing $4$ to $1$ persons given $0$ children and using livebait, the expected difference is about $26$ fish. When not using livebait, the expected difference decreases substantially to about $5$ fish. Using livebait with a group of people is associated with a much larger expected difference in the number of fish caught. \n", + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fig, ax = bmb.interpret.plot_comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=[\"persons\", \"child\"],\n", + " subplot_kwargs={\"main\": \"persons\", \"group\": \"child\", \"panel\": \"child\"},\n", + " fig_kwargs={\"figsize\":(12, 3), \"sharey\": True},\n", + " legend=False\n", + ") " + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Unit level contrasts\n", + "\n", + "Evaluating average predictive comparisons at central values for the conditional covariates $c$ can be problematic when the inputs have a large variance since no single central value (mean, median, etc.) is representative of the covariate. This is especially true when $c$ exhibits bi or multimodality. Thus, it may be desireable to use the empirical distribution of $c$ to compute the predictive comparisons, and then average over a specific or set of covariates to obtain the average predictive comparisons. To achieve unit level contrasts, do not pass a parameter into `conditional` and or specify `None`." + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "True\n" + ] + }, + { + "data": { + "text/html": [ + "
\n", + "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevaluechildlivebaitcamperestimatelower_3.0%upper_97.0%
0personsdiff(1.0, 4.0)0.00.01.04.8344722.5634727.037150
1personsdiff(1.0, 4.0)0.01.01.026.42318823.73972929.072748
2personsdiff(1.0, 4.0)1.00.01.01.2020030.6316291.780965
3personsdiff(1.0, 4.0)1.01.01.06.5719435.4692757.642248
4personsdiff(1.0, 4.0)2.00.01.00.3013840.1436760.467608
5personsdiff(1.0, 4.0)2.01.01.01.6484171.1404152.187190
\n", - "
" - ], - "text/plain": [ - " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", - "0 persons diff (1.0, 4.0) ... 4.834472 2.563472 7.037150\n", - "1 persons diff (1.0, 4.0) ... 26.423188 23.739729 29.072748\n", - "2 persons diff (1.0, 4.0) ... 1.202003 0.631629 1.780965\n", - "3 persons diff (1.0, 4.0) ... 6.571943 5.469275 7.642248\n", - "4 persons diff (1.0, 4.0) ... 0.301384 0.143676 0.467608\n", - "5 persons diff (1.0, 4.0) ... 1.648417 1.140415 2.187190\n", - "\n", - "[6 rows x 9 columns]" - ] - }, - "execution_count": 5, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast={\"persons\": [1, 4]},\n", - " conditional={\"child\": [0, 1, 2], \"livebait\": [0, 1]},\n", - ") " - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "But why is `camper` also in the summary dataframe? This is because in order to peform predictions, Bambi is expecting a value for each covariate used to fit the model. Additionally, with GLM models, average predictive comparisons are conditional in the sense that the estimate depends on the values of all the covariates in the model. Thus, for unspecified covariates, `comparisons` and `plot_comparisons` computes a default value (mean or mode based on the data type of the covariate). Thus, $c$ = `child`, `livebait`, `camper`. Each row in the summary dataframe is read as \"comparing $4$ to $1$ persons conditional on $c$, the expected difference in the outcome is $y$.\"" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### Multiple contrast values\n", + " .dataframe tbody tr th {\n", + " vertical-align: top;\n", + " }\n", "\n", - "Users can also perform comparisons on multiple contrast values. For example, if we wanted to compare the number of fish caught between $(1, 2)$, $(1, 4)$, and $(2, 4)$ `persons` conditional on a range of values for `child` and `livebait`." - ] - }, - { - "cell_type": "code", - "execution_count": 6, - "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "
\n", - "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevaluechildlivebaitcamperestimatelower_3.0%upper_97.0%
0personsdiff(1, 2)0.00.01.00.5276270.2954510.775465
1personsdiff(1, 2)0.01.01.02.8836942.6056903.177685
2personsdiff(1, 2)1.00.01.00.1313190.0673390.195132
3personsdiff(1, 2)1.01.01.00.7179650.5929680.857893
4personsdiff(1, 2)2.00.01.00.0329600.0152120.052075
5personsdiff(1, 2)2.01.01.00.1802700.1231730.244695
6personsdiff(1, 4)0.00.01.04.8344722.5634727.037150
7personsdiff(1, 4)0.01.01.026.42318823.73972929.072748
8personsdiff(1, 4)1.00.01.01.2020030.6316291.780965
9personsdiff(1, 4)1.01.01.06.5719435.4692757.642248
10personsdiff(1, 4)2.00.01.00.3013840.1436760.467608
11personsdiff(1, 4)2.01.01.01.6484171.1404152.187190
12personsdiff(2, 4)0.00.01.04.3068452.2670976.280005
13personsdiff(2, 4)0.01.01.023.53949420.99093126.240169
14personsdiff(2, 4)1.00.01.01.0706830.5659311.585718
15personsdiff(2, 4)1.01.01.05.8539784.8589576.848519
16personsdiff(2, 4)2.00.01.00.2684230.1240330.412274
17personsdiff(2, 4)2.01.01.01.4681471.0248001.960934
\n", - "
" - ], - "text/plain": [ - " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", - "0 persons diff (1, 2) ... 0.527627 0.295451 0.775465\n", - "1 persons diff (1, 2) ... 2.883694 2.605690 3.177685\n", - "2 persons diff (1, 2) ... 0.131319 0.067339 0.195132\n", - "3 persons diff (1, 2) ... 0.717965 0.592968 0.857893\n", - "4 persons diff (1, 2) ... 0.032960 0.015212 0.052075\n", - "5 persons diff (1, 2) ... 0.180270 0.123173 0.244695\n", - "6 persons diff (1, 4) ... 4.834472 2.563472 7.037150\n", - "7 persons diff (1, 4) ... 26.423188 23.739729 29.072748\n", - "8 persons diff (1, 4) ... 1.202003 0.631629 1.780965\n", - "9 persons diff (1, 4) ... 6.571943 5.469275 7.642248\n", - "10 persons diff (1, 4) ... 0.301384 0.143676 0.467608\n", - "11 persons diff (1, 4) ... 1.648417 1.140415 2.187190\n", - "12 persons diff (2, 4) ... 4.306845 2.267097 6.280005\n", - "13 persons diff (2, 4) ... 23.539494 20.990931 26.240169\n", - "14 persons diff (2, 4) ... 1.070683 0.565931 1.585718\n", - "15 persons diff (2, 4) ... 5.853978 4.858957 6.848519\n", - "16 persons diff (2, 4) ... 0.268423 0.124033 0.412274\n", - "17 persons diff (2, 4) ... 1.468147 1.024800 1.960934\n", - "\n", - "[18 rows x 9 columns]" - ] - }, - "execution_count": 6, - "metadata": {}, - "output_type": "execute_result" - } + " .dataframe thead th {\n", + " text-align: right;\n", + " }\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevaluecamperchildpersonsestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)0.00.01.00.8644080.6270631.116105
1livebaitdiff(0.0, 1.0)1.00.01.01.6946461.2528032.081207
2livebaitdiff(0.0, 1.0)0.00.01.00.8644080.6270631.116105
3livebaitdiff(0.0, 1.0)1.01.02.01.0090940.7554491.249551
4livebaitdiff(0.0, 1.0)0.00.01.00.8644080.6270631.116105
5livebaitdiff(0.0, 1.0)1.02.04.01.4532350.9646741.956434
6livebaitdiff(0.0, 1.0)0.01.03.01.2332470.9002951.569891
7livebaitdiff(0.0, 1.0)0.03.04.00.1880190.0903280.289560
8livebaitdiff(0.0, 1.0)1.02.03.00.6063610.3905710.818549
9livebaitdiff(0.0, 1.0)1.00.01.01.6946461.2528032.081207
\n", + "" ], - "source": [ - "multiple_values = comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast={\"persons\": [1, 2, 4]},\n", - " conditional={\"child\": [0, 1, 2], \"livebait\": [0, 1]}\n", - ")\n", + "text/plain": [ + " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", + "0 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", + "1 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", + "2 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", + "3 livebait diff (0.0, 1.0) ... 1.009094 0.755449 1.249551\n", + "4 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", + "5 livebait diff (0.0, 1.0) ... 1.453235 0.964674 1.956434\n", + "6 livebait diff (0.0, 1.0) ... 1.233247 0.900295 1.569891\n", + "7 livebait diff (0.0, 1.0) ... 0.188019 0.090328 0.289560\n", + "8 livebait diff (0.0, 1.0) ... 0.606361 0.390571 0.818549\n", + "9 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", "\n", - "multiple_values" + "[10 rows x 9 columns]" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "Notice how the contrast $w$ varies while the covariates $c$ are held constant. Currently, however, plotting multiple contrast values can be difficult to interpret since the contrast is \"abstracted\" away onto the y-axis. Thus, it would be difficult to interpret which portion of the plot corresponds to which contrast value. Therefore, it is currently recommended that if you want to plot multiple contrast values, call `comparisons` directly to obtain the summary dataframe and plot the results yourself." - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### Default contrast and conditional values\n", - "\n", - "Now, we move onto scenario 2 described above (grid of equally spaced and central values) in computing average predictive comparisons. You are not required to pass values for `contrast` and `conditional`. If you do not pass values, Bambi will compute default values for you. Below, it is described how these default values are computed.\n", + "execution_count": 10, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "unit_level = bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=None,\n", + ")\n", + "\n", + "# empirical distribution\n", + "print(unit_level.shape[0] == fish_model.data.shape[0])\n", + "unit_level.head(10)" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevaluepersonschildcamperestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)1.0000000.01.01.6946461.2528032.081207
1livebaitdiff(0.0, 1.0)1.0000001.01.00.4224480.2990520.551766
2livebaitdiff(0.0, 1.0)1.0000003.01.00.0269230.0127520.043035
3livebaitdiff(0.0, 1.0)1.0612240.01.01.7874121.3429792.203158
4livebaitdiff(0.0, 1.0)1.0612241.01.00.4455550.3172530.580117
5livebaitdiff(0.0, 1.0)1.0612243.01.00.0283930.0134520.045276
6livebaitdiff(0.0, 1.0)1.1224490.01.01.8852701.4229372.313218
7livebaitdiff(0.0, 1.0)1.1224491.01.00.4699290.3353730.609249
8livebaitdiff(0.0, 1.0)1.1224493.01.00.0299440.0141650.047593
9livebaitdiff(0.0, 1.0)1.1836740.01.01.9885001.5016502.424762
\n", - "
" - ], - "text/plain": [ - " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", - "0 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", - "1 livebait diff (0.0, 1.0) ... 0.422448 0.299052 0.551766\n", - "2 livebait diff (0.0, 1.0) ... 0.026923 0.012752 0.043035\n", - "3 livebait diff (0.0, 1.0) ... 1.787412 1.342979 2.203158\n", - "4 livebait diff (0.0, 1.0) ... 0.445555 0.317253 0.580117\n", - "5 livebait diff (0.0, 1.0) ... 0.028393 0.013452 0.045276\n", - "6 livebait diff (0.0, 1.0) ... 1.885270 1.422937 2.313218\n", - "7 livebait diff (0.0, 1.0) ... 0.469929 0.335373 0.609249\n", - "8 livebait diff (0.0, 1.0) ... 0.029944 0.014165 0.047593\n", - "9 livebait diff (0.0, 1.0) ... 1.988500 1.501650 2.424762\n", - "\n", - "[10 rows x 9 columns]" - ] - }, - "execution_count": 7, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "contrast_df = comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=[\"persons\", \"child\"],\n", - ")\n", - "\n", - "contrast_df.head(10)" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "As `livebait` was encoded as a categorical dtype, Bambi returned the unique levels of $[0, 1]$ for the contrast. `persons` and `child` were passed as the first and second element and thus act as the main and group variables, respectively. It can be see from the output above, that an equally spaced grid was used to compute the values for `persons`, whereas a quantile based grid was used for `child`. Furthermore, as `camper` was unspecified, the mode was used as the default value. Lets go ahead and plot the commparisons." - ] - }, - { - "cell_type": "code", - "execution_count": 8, - "metadata": {}, - "outputs": [ - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "fig, ax = plot_comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=[\"persons\", \"child\"],\n", - ") \n", - "fig.set_size_inches(7, 3)" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "The plot shows us that the expected differences in fish caught comparing a group of people who use livebait and no livebait is not only conditional on the number of persons, but also children. However, the plotted comparisons for `child` = $3$ is difficult to interpret on a single plot. Thus, it can be useful to pass specific `group` and `panel` arguments to aid in the interpretation of the plot. Therefore, `subplot_kwargs` allows the user to manipulate the plotting by passing a dictionary where the keys are `{\"main\": ..., \"group\": ..., \"panel\": ...}` and the values are the names of the covariates to be plotted. Below, we plot the same comparisons as above, but this time we specify `group` and `panel` to both be `child`." - ] - }, - { - "cell_type": "code", - "execution_count": 9, - "metadata": {}, - "outputs": [ - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } + " .dataframe thead th {\n", + " text-align: right;\n", + " }\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
countlivebaitcamperpersonschild
00.00.00.01.00.0
10.01.01.01.00.0
20.01.00.01.00.0
30.01.01.02.01.0
41.01.00.01.00.0
50.01.01.04.02.0
60.01.00.03.01.0
70.01.00.04.03.0
80.00.01.03.02.0
91.01.01.01.00.0
\n", + "" ], - "source": [ - "fig, ax = plot_comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=[\"persons\", \"child\"],\n", - " subplot_kwargs={\"main\": \"persons\", \"group\": \"child\", \"panel\": \"child\"},\n", - " fig_kwargs={\"figsize\":(12, 3), \"sharey\": True},\n", - " legend=False\n", - ") " + "text/plain": [ + " count livebait camper persons child\n", + "0 0.0 0.0 0.0 1.0 0.0\n", + "1 0.0 1.0 1.0 1.0 0.0\n", + "2 0.0 1.0 0.0 1.0 0.0\n", + "3 0.0 1.0 1.0 2.0 1.0\n", + "4 1.0 1.0 0.0 1.0 0.0\n", + "5 0.0 1.0 1.0 4.0 2.0\n", + "6 0.0 1.0 0.0 3.0 1.0\n", + "7 0.0 1.0 0.0 4.0 3.0\n", + "8 0.0 0.0 1.0 3.0 2.0\n", + "9 1.0 1.0 1.0 1.0 0.0" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### Unit level contrasts\n", + "execution_count": 11, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# empirical (observed) data used to fit the model\n", + "fish_model.data.head(10)" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Above, `unit_level` is the comparisons summary dataframe and `fish_model.data` is the empirical data. Notice how the values for $c$ are identical in both dataframes. However, for $w$, the values are different. However, these unit level contrasts are difficult to interpret as each row corresponds to _that_ unit's contrast. Therefore, it is useful to average over (marginalize) the estimates to summarize the unit level predictive comparisons." + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "#### Marginalizing over covariates" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Since the empirical distrubution is used for computing the average predictive comparisons, the same number of rows (250) is returned as the data used to fit the model. To average over a covariate, use the `average_by` argument. If `True` is passed, then `comparisons` averages over all covariates. Else, if a single or list of covariates are passed, then `comparisons` averages by the covariates passed." + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevaluecamperchildpersonsestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)0.00.01.00.8644080.6270631.116105
1livebaitdiff(0.0, 1.0)1.00.01.01.6946461.2528032.081207
2livebaitdiff(0.0, 1.0)0.00.01.00.8644080.6270631.116105
3livebaitdiff(0.0, 1.0)1.01.02.01.0090940.7554491.249551
4livebaitdiff(0.0, 1.0)0.00.01.00.8644080.6270631.116105
5livebaitdiff(0.0, 1.0)1.02.04.01.4532350.9646741.956434
6livebaitdiff(0.0, 1.0)0.01.03.01.2332470.9002951.569891
7livebaitdiff(0.0, 1.0)0.03.04.00.1880190.0903280.289560
8livebaitdiff(0.0, 1.0)1.02.03.00.6063610.3905710.818549
9livebaitdiff(0.0, 1.0)1.00.01.01.6946461.2528032.081207
\n", - "
" - ], - "text/plain": [ - " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", - "0 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", - "1 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", - "2 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", - "3 livebait diff (0.0, 1.0) ... 1.009094 0.755449 1.249551\n", - "4 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", - "5 livebait diff (0.0, 1.0) ... 1.453235 0.964674 1.956434\n", - "6 livebait diff (0.0, 1.0) ... 1.233247 0.900295 1.569891\n", - "7 livebait diff (0.0, 1.0) ... 0.188019 0.090328 0.289560\n", - "8 livebait diff (0.0, 1.0) ... 0.606361 0.390571 0.818549\n", - "9 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", - "\n", - "[10 rows x 9 columns]" - ] - }, - "execution_count": 10, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "unit_level = comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=None,\n", - ")\n", + " .dataframe tbody tr th {\n", + " vertical-align: top;\n", + " }\n", "\n", - "# empirical distribution\n", - "print(unit_level.shape[0] == fish_model.data.shape[0])\n", - "unit_level.head(10)" - ] - }, - { - "cell_type": "code", - "execution_count": 11, - "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "
\n", - "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
countlivebaitcamperpersonschild
00.00.00.01.00.0
10.01.01.01.00.0
20.01.00.01.00.0
30.01.01.02.01.0
41.01.00.01.00.0
50.01.01.04.02.0
60.01.00.03.01.0
70.01.00.04.03.0
80.00.01.03.02.0
91.01.01.01.00.0
\n", - "
" - ], - "text/plain": [ - " count livebait camper persons child\n", - "0 0.0 0.0 0.0 1.0 0.0\n", - "1 0.0 1.0 1.0 1.0 0.0\n", - "2 0.0 1.0 0.0 1.0 0.0\n", - "3 0.0 1.0 1.0 2.0 1.0\n", - "4 1.0 1.0 0.0 1.0 0.0\n", - "5 0.0 1.0 1.0 4.0 2.0\n", - "6 0.0 1.0 0.0 3.0 1.0\n", - "7 0.0 1.0 0.0 4.0 3.0\n", - "8 0.0 0.0 1.0 3.0 2.0\n", - "9 1.0 1.0 1.0 1.0 0.0" - ] - }, - "execution_count": 11, - "metadata": {}, - "output_type": "execute_result" - } + " .dataframe thead th {\n", + " text-align: right;\n", + " }\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevalueestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)3.6496912.9561854.333621
\n", + "" ], - "source": [ - "# empirical (observed) data used to fit the model\n", - "fish_model.data.head(10)" + "text/plain": [ + " term estimate_type value estimate lower_3.0% upper_97.0%\n", + "0 livebait diff (0.0, 1.0) 3.649691 2.956185 4.333621" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "Above, `unit_level` is the comparisons summary dataframe and `fish_model.data` is the empirical data. Notice how the values for $c$ are identical in both dataframes. However, for $w$, the values are different. However, these unit level contrasts are difficult to interpret as each row corresponds to _that_ unit's contrast. Therefore, it is useful to average over (marginalize) the estimates to summarize the unit level predictive comparisons." + "execution_count": 12, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# marginalize over all covariates\n", + "bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=None,\n", + " average_by=True\n", + ")" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Passing `True` to `average_by` averages over all covariates and is equivalent to taking the mean of the `estimate` and uncertainty columns. For example:" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "estimate 3.649691\n", + "lower_3.0% 2.956185\n", + "upper_97.0% 4.333621\n", + "dtype: float64" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "#### Marginalizing over covariates" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "Since the empirical distrubution is used for computing the average predictive comparisons, the same number of rows (250) is returned as the data used to fit the model. To average over a covariate, use the `average_by` argument. If `True` is passed, then `comparisons` averages over all covariates. Else, if a single or list of covariates are passed, then `comparisons` averages by the covariates passed." - ] - }, - { - "cell_type": "code", - "execution_count": 12, - "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "
\n", - "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevalueestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)3.6496912.9561854.333621
\n", - "
" - ], - "text/plain": [ - " term estimate_type value estimate lower_3.0% upper_97.0%\n", - "0 livebait diff (0.0, 1.0) 3.649691 2.956185 4.333621" - ] - }, - "execution_count": 12, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "# marginalize over all covariates\n", - "comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=None,\n", - " average_by=True\n", - ")" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "Passing `True` to `average_by` averages over all covariates and is equivalent to taking the mean of the `estimate` and uncertainty columns. For example:" - ] - }, - { - "cell_type": "code", - "execution_count": 13, - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "estimate 3.649691\n", - "lower_3.0% 2.956185\n", - "upper_97.0% 4.333621\n", - "dtype: float64" - ] - }, - "execution_count": 13, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "unit_level = comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=None,\n", - ")\n", + "execution_count": 13, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "unit_level = bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=None,\n", + ")\n", + "\n", + "unit_level[[\"estimate\", \"lower_3.0%\", \"upper_97.0%\"]].mean()" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "#### Average by subgroups\n", + "\n", + "Averaging over all covariates may not be desired, and you would rather average by a group or specific covariate. To perform averaging by subgroups, users can pass a single or list of covariates to `average_by` to average over specific covariates. For example, if we wanted to average by `persons`:" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevaluepersonsestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)1.01.3742031.0112901.708711
1livebaitdiff(0.0, 1.0)2.01.9633621.5433302.376636
2livebaitdiff(0.0, 1.0)3.03.7015103.0565864.357385
3livebaitdiff(0.0, 1.0)4.07.3586626.0476428.655654
\n", - "
" - ], - "text/plain": [ - " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", - "0 livebait diff (0.0, 1.0) ... 1.374203 1.011290 1.708711\n", - "1 livebait diff (0.0, 1.0) ... 1.963362 1.543330 2.376636\n", - "2 livebait diff (0.0, 1.0) ... 3.701510 3.056586 4.357385\n", - "3 livebait diff (0.0, 1.0) ... 7.358662 6.047642 8.655654\n", - "\n", - "[4 rows x 7 columns]" - ] - }, - "execution_count": 14, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "# average by number of persons\n", - "comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=None,\n", - " average_by=\"persons\"\n", - ")" - ] - }, - { - "cell_type": "code", - "execution_count": 15, - "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "
\n", - "\n", - "\n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - " \n", - "
termestimate_typevaluepersonscamperestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)1.00.00.8644080.6270631.116105
1livebaitdiff(0.0, 1.0)1.01.01.6946461.2528032.081207
2livebaitdiff(0.0, 1.0)2.00.01.4245981.0783891.777154
3livebaitdiff(0.0, 1.0)2.01.02.3444391.8721912.800661
4livebaitdiff(0.0, 1.0)3.00.02.4294591.8715782.964242
5livebaitdiff(0.0, 1.0)3.01.04.4435403.7478405.170052
6livebaitdiff(0.0, 1.0)4.00.03.5419212.6864454.391176
7livebaitdiff(0.0, 1.0)4.01.010.7392049.02470212.432764
\n", - "
" - ], - "text/plain": [ - " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", - "0 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", - "1 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", - "2 livebait diff (0.0, 1.0) ... 1.424598 1.078389 1.777154\n", - "3 livebait diff (0.0, 1.0) ... 2.344439 1.872191 2.800661\n", - "4 livebait diff (0.0, 1.0) ... 2.429459 1.871578 2.964242\n", - "5 livebait diff (0.0, 1.0) ... 4.443540 3.747840 5.170052\n", - "6 livebait diff (0.0, 1.0) ... 3.541921 2.686445 4.391176\n", - "7 livebait diff (0.0, 1.0) ... 10.739204 9.024702 12.432764\n", - "\n", - "[8 rows x 8 columns]" - ] - }, - "execution_count": 15, - "metadata": {}, - "output_type": "execute_result" - } + " .dataframe thead th {\n", + " text-align: right;\n", + " }\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevaluepersonsestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)1.01.3742031.0112901.708711
1livebaitdiff(0.0, 1.0)2.01.9633621.5433302.376636
2livebaitdiff(0.0, 1.0)3.03.7015103.0565864.357385
3livebaitdiff(0.0, 1.0)4.07.3586626.0476428.655654
\n", + "" ], - "source": [ - "# average by number of persons and camper by passing a list\n", - "comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=None,\n", - " average_by=[\"persons\", \"camper\"]\n", - ")" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "It is still possible to use `plot_comparisons` when passing an argument to `average_by`. In the plot below, the empirical distribution is used to compute unit level contrasts for `livebait` and then averaged over `persons` to obtain the average predictive comparisons. The plot below is similar to the second plot in this notebook. The differences being that: (1) a pairwise transition grid is defined for the second plot above, whereas the empirical distribution is used in the plot below, and (2) in the plot below, we marginalized over the other covariates in the model (thus the reason for not having a `camper` or `child` group and panel, and a reduction in the uncertainty interval)." - ] - }, - { - "cell_type": "code", - "execution_count": 16, - "metadata": {}, - "outputs": [ - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "fig, ax = plot_comparisons(\n", - " model=fish_model,\n", - " idata=fish_idata,\n", - " contrast=\"livebait\",\n", - " conditional=None,\n", - " average_by=\"persons\"\n", - ")\n", - "fig.set_size_inches(7, 3)" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "## Logistic Regression\n", + "text/plain": [ + " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", + "0 livebait diff (0.0, 1.0) ... 1.374203 1.011290 1.708711\n", + "1 livebait diff (0.0, 1.0) ... 1.963362 1.543330 2.376636\n", + "2 livebait diff (0.0, 1.0) ... 3.701510 3.056586 4.357385\n", + "3 livebait diff (0.0, 1.0) ... 7.358662 6.047642 8.655654\n", "\n", - "To showcase an additional functionality of `comparisons` and `plot_comparisons`, we fit a logistic regression model to the [titanic dataset](https://vincentarelbundock.github.io/Rdatasets/csv/Stat2Data/Titanic.csv) with interaction terms to model the probability of survival. The titanic dataset gives the values of four categorical attributes for each of the 2201 people on board the Titanic when it struck an iceberg and sank. The attributes are social class (first class, second class, third class, crewmember), age, sex (0 = female, 1 = male), and whether or not the person survived (0 = deceased, 1 = survived). " + "[4 rows x 7 columns]" ] }, - { - "cell_type": "code", - "execution_count": 17, - "metadata": {}, - "outputs": [], - "source": [ - "dat = pd.read_csv(\"https://vincentarelbundock.github.io/Rdatasets/csv/Stat2Data/Titanic.csv\", index_col=0)\n", + "execution_count": 14, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# average by number of persons\n", + "bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=None,\n", + " average_by=\"persons\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
termestimate_typevaluepersonscamperestimatelower_3.0%upper_97.0%
0livebaitdiff(0.0, 1.0)1.00.00.8644080.6270631.116105
1livebaitdiff(0.0, 1.0)1.01.01.6946461.2528032.081207
2livebaitdiff(0.0, 1.0)2.00.01.4245981.0783891.777154
3livebaitdiff(0.0, 1.0)2.01.02.3444391.8721912.800661
4livebaitdiff(0.0, 1.0)3.00.02.4294591.8715782.964242
5livebaitdiff(0.0, 1.0)3.01.04.4435403.7478405.170052
6livebaitdiff(0.0, 1.0)4.00.03.5419212.6864454.391176
7livebaitdiff(0.0, 1.0)4.01.010.7392049.02470212.432764
\n", + "
" ], - "source": [ - "titanic_model = bmb.Model(\n", - " \"Survived ~ PClass * SexCode * Age\", \n", - " data=dat, \n", - " family=\"bernoulli\"\n", - ")\n", - "titanic_idata = titanic_model.fit(draws=1000, target_accept=0.95, random_seed=1234)" - ] - }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### Comparison types\n", + "text/plain": [ + " term estimate_type value ... estimate lower_3.0% upper_97.0%\n", + "0 livebait diff (0.0, 1.0) ... 0.864408 0.627063 1.116105\n", + "1 livebait diff (0.0, 1.0) ... 1.694646 1.252803 2.081207\n", + "2 livebait diff (0.0, 1.0) ... 1.424598 1.078389 1.777154\n", + "3 livebait diff (0.0, 1.0) ... 2.344439 1.872191 2.800661\n", + "4 livebait diff (0.0, 1.0) ... 2.429459 1.871578 2.964242\n", + "5 livebait diff (0.0, 1.0) ... 4.443540 3.747840 5.170052\n", + "6 livebait diff (0.0, 1.0) ... 3.541921 2.686445 4.391176\n", + "7 livebait diff (0.0, 1.0) ... 10.739204 9.024702 12.432764\n", "\n", - "`comparisons` and `plot_comparisons` also allow you to specify the type of comparison to be computed. By default, a difference is used. However, it is also possible to take the ratio where comparisons would then become _average predictive ratios_. To achieve this, pass `\"ratio\"` into the argument `comparison_type`. Using different comparison types offers a way to produce alternative insights; especially when there are interaction terms as the value of one covariate depends on the value of the other covariate." + "[8 rows x 8 columns]" ] }, - { - "cell_type": "code", - "execution_count": 19, - "metadata": {}, - "outputs": [ - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "fig, ax = plot_comparisons(\n", - " model=titanic_model,\n", - " idata=titanic_idata,\n", - " contrast={\"PClass\": [1, 3]},\n", - " conditional=[\"Age\", \"SexCode\"],\n", - " comparison_type=\"ratio\",\n", - " subplot_kwargs={\"main\": \"Age\", \"group\": \"SexCode\", \"panel\": \"SexCode\"},\n", - " fig_kwargs={\"figsize\":(12, 3), \"sharey\": True},\n", - " legend=False\n", - "\n", - ")" + "execution_count": 15, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# average by number of persons and camper by passing a list\n", + "bmb.interpret.comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=None,\n", + " average_by=[\"persons\", \"camper\"]\n", + ")" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "It is still possible to use `plot_comparisons` when passing an argument to `average_by`. In the plot below, the empirical distribution is used to compute unit level contrasts for `livebait` and then averaged over `persons` to obtain the average predictive comparisons. The plot below is similar to the second plot in this notebook. The differences being that: (1) a pairwise transition grid is defined for the second plot above, whereas the empirical distribution is used in the plot below, and (2) in the plot below, we marginalized over the other covariates in the model (thus the reason for not having a `camper` or `child` group and panel, and a reduction in the uncertainty interval)." + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" ] }, - { - "attachments": {}, - "cell_type": "markdown", - "metadata": {}, - "source": [ - "The left panel shows that the ratio of the probability of survival comparing `PClass` $3$ to $1$ conditional on `Age` is non-constant. Whereas the right panel shows an approximately constant ratio in the probability of survival comparing `PClass` $3$ to $1$ conditional on `Age`. " + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fig, ax = bmb.interpret.plot_comparisons(\n", + " model=fish_model,\n", + " idata=fish_idata,\n", + " contrast=\"livebait\",\n", + " conditional=None,\n", + " average_by=\"persons\"\n", + ")\n", + "fig.set_size_inches(7, 3)" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Logistic Regression\n", + "\n", + "To showcase an additional functionality of `comparisons` and `plot_comparisons`, we fit a logistic regression model to the [titanic dataset](https://vincentarelbundock.github.io/Rdatasets/csv/Stat2Data/Titanic.csv) with interaction terms to model the probability of survival. The titanic dataset gives the values of four categorical attributes for each of the 2201 people on board the Titanic when it struck an iceberg and sank. The attributes are social class (first class, second class, third class, crewmember), age, sex (0 = female, 1 = male), and whether or not the person survived (0 = deceased, 1 = survived). " + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": {}, + "outputs": [], + "source": [ + "dat = pd.read_csv(\"https://vincentarelbundock.github.io/Rdatasets/csv/Stat2Data/Titanic.csv\", index_col=0)\n", + "\n", + "dat[\"PClass\"] = dat[\"PClass\"].str.replace(\"[st, nd, rd]\", \"\", regex=True)\n", + "dat[\"PClass\"] = dat[\"PClass\"].str.replace(\"*\", \"0\").astype(int)\n", + "dat[\"PClass\"] = dat[\"PClass\"].replace(0, np.nan)\n", + "dat[\"PClass\"] = pd.Categorical(dat[\"PClass\"], ordered=True)\n", + "dat[\"SexCode\"] = pd.Categorical(dat[\"SexCode\"], ordered=True)\n", + "\n", + "dat = dat.dropna(axis=0, how=\"any\")" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Modeling the probability that Survived==1\n", + "Auto-assigning NUTS sampler...\n", + "Initializing NUTS using jitter+adapt_diag...\n", + "Multiprocess sampling (4 chains in 4 jobs)\n", + "NUTS: [Intercept, PClass, SexCode, PClass:SexCode, Age, PClass:Age, SexCode:Age, PClass:SexCode:Age]\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + " |████████████████████████████████| 100.00% [8000/8000 00:15<00:00 Sampling 4 chains, 0 divergences]\r" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 16 seconds.\n" + ] + } + ], + "source": [ + "titanic_model = bmb.Model(\n", + " \"Survived ~ PClass * SexCode * Age\", \n", + " data=dat, \n", + " family=\"bernoulli\"\n", + ")\n", + "titanic_idata = titanic_model.fit(draws=1000, target_accept=0.95, random_seed=1234)" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Comparison types\n", + "\n", + "`comparisons` and `plot_comparisons` also allow you to specify the type of comparison to be computed. By default, a difference is used. However, it is also possible to take the ratio where comparisons would then become _average predictive ratios_. To achieve this, pass `\"ratio\"` into the argument `comparison_type`. Using different comparison types offers a way to produce alternative insights; especially when there are interaction terms as the value of one covariate depends on the value of the other covariate." + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "metadata": {}, + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "
" ] - } - ], - "metadata": { - "kernelspec": { - "display_name": "bambinos", - "language": "python", - "name": "python3" }, - "language_info": { - "codemirror_mode": { - "name": "ipython", - "version": 3 - }, - "file_extension": ".py", - "mimetype": "text/x-python", - "name": "python", - "nbconvert_exporter": "python", - "pygments_lexer": "ipython3", - "version": "3.11.0" - }, - "orig_nbformat": 4 - }, - "nbformat": 4, - "nbformat_minor": 2 - } \ No newline at end of file + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fig, ax = bmb.interpret.plot_comparisons(\n", + " model=titanic_model,\n", + " idata=titanic_idata,\n", + " contrast={\"PClass\": [1, 3]},\n", + " conditional=[\"Age\", \"SexCode\"],\n", + " comparison_type=\"ratio\",\n", + " subplot_kwargs={\"main\": \"Age\", \"group\": \"SexCode\", \"panel\": \"SexCode\"},\n", + " fig_kwargs={\"figsize\":(12, 3), \"sharey\": True},\n", + " legend=False\n", + "\n", + ")" + ] + }, + { + "attachments": {}, + "cell_type": "markdown", + "metadata": {}, + "source": [ + "The left panel shows that the ratio of the probability of survival comparing `PClass` $3$ to $1$ conditional on `Age` is non-constant. Whereas the right panel shows an approximately constant ratio in the probability of survival comparing `PClass` $3$ to $1$ conditional on `Age`. " + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Last updated: Wed Nov 01 2023\n", + "\n", + "Python implementation: CPython\n", + "Python version : 3.11.0\n", + "IPython version : 8.13.2\n", + "\n", + "bambi : 0.13.0.dev0\n", + "pandas: 2.1.0\n", + "numpy : 1.24.2\n", + "arviz : 0.16.1\n", + "\n", + "Watermark: 2.3.1\n", + "\n" + ] + } + ], + "source": [ + "%load_ext watermark\n", + "%watermark -n -u -v -iv -w" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "bambinos", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.0" + }, + "orig_nbformat": 4 + }, + "nbformat": 4, + "nbformat_minor": 2 +}