Update slides 11

content/MATH80667A-10-content.R

@@ -92,11 +92,14 @@ head(C22)
 options(contrasts = c("contr.sum", "contr.poly"))
 # balanced!
 xtabs(~ anchor + vignette + verdictsyst, data = C22)
+# Incomplete block design: there are two (counterbalanced) combo of anchor, vignette, verdictsyst
+# but participants see 2 out of 8 combinations. So no interaction between these and ID is possible.
+xtabs(~  id + interaction(anchor, vignette, verdictsyst), data = C22)
 model <- lmer(
-  guilt ~ anchor*vignette*verdictsyst + prior + (1|id),
+  guilt ~ anchor*vignette*verdictsyst + pjaq + (1|id),
   data = C22)
-# prior is a covariate (so used to reduce error, plus the slope is of interest on it's own
-# Cannot have interaction prior * id, because we get a single prior score per person
+# pjaq is a covariate (so used to reduce error, plus the slope is of interest on it's own
+# Cannot have interaction pjaq * id, because we get a single pjaq score per person
 
 # No ambiguity for sum of square decomposition
 anova(model)

content/MATH80667A-10-content.sps

@@ -1,9 +1,23 @@
+GET FILE='C:\Users\11187439\Desktop\SPSS\workhours.sav'. 
+DATASET NAME workhours WINDOW=FRONT.
+
+MIXED time
+  /CRITERIA=DFMETHOD(SATTERTHWAITE)
+  /FIXED= | SSTYPE(3) 
+  /METHOD=REML 
+  /PRINT=SOLUTION 
+  /RANDOM=INTERCEPT | SUBJECT(id) COVTYPE(VC).
+
+
+DATASET CLOSE workhours.
+
+
 GET FILE='C:\Users\11187439\Desktop\SPSS\C22_E3.sav'. 
 DATASET NAME C22 WINDOW=FRONT.
 
-MIXED guilt BY anchor vignette verdictsyst WITH prior
+MIXED guilt BY anchor vignette verdictsyst WITH pjaq
   /CRITERIA=DFMETHOD(SATTERTHWAITE)
-  /FIXED=anchor vignette verdictsyst anchor*vignette anchor*verdictsyst vignette*verdictsyst anchor*vignette*verdictsyst prior | SSTYPE(3) 
+  /FIXED=anchor vignette verdictsyst anchor*vignette anchor*verdictsyst vignette*verdictsyst anchor*vignette*verdictsyst pjaq | SSTYPE(3) 
   /METHOD=REML 
   /PRINT=SOLUTION 
   /RANDOM=INTERCEPT | SUBJECT(id) COVTYPE(VC).
@@ -23,4 +37,4 @@ MIXED y BY choc background
   /RANDOM=INTERCEPT choc | SUBJECT(rater*background) COVTYPE(VC).
 
 
-DATASET CLOSE chocolate.
+DATASET CLOSE chocolate.

docs/content/MATH80667A-10-content.R

@@ -92,11 +92,14 @@ head(C22)
 options(contrasts = c("contr.sum", "contr.poly"))
 # balanced!
 xtabs(~ anchor + vignette + verdictsyst, data = C22)
+# Incomplete block design: there are two (counterbalanced) combo of anchor, vignette, verdictsyst
+# but participants see 2 out of 8 combinations. So no interaction between these and ID is possible.
+xtabs(~  id + interaction(anchor, vignette, verdictsyst), data = C22)
 model <- lmer(
-  guilt ~ anchor*vignette*verdictsyst + prior + (1|id),
+  guilt ~ anchor*vignette*verdictsyst + pjaq + (1|id),
   data = C22)
-# prior is a covariate (so used to reduce error, plus the slope is of interest on it's own
-# Cannot have interaction prior * id, because we get a single prior score per person
+# pjaq is a covariate (so used to reduce error, plus the slope is of interest on it's own
+# Cannot have interaction pjaq * id, because we get a single pjaq score per person
 
 # No ambiguity for sum of square decomposition
 anova(model)

docs/content/MATH80667A-10-content.sps

@@ -1,9 +1,23 @@
+GET FILE='C:\Users\11187439\Desktop\SPSS\workhours.sav'. 
+DATASET NAME workhours WINDOW=FRONT.
+
+MIXED time
+  /CRITERIA=DFMETHOD(SATTERTHWAITE)
+  /FIXED= | SSTYPE(3) 
+  /METHOD=REML 
+  /PRINT=SOLUTION 
+  /RANDOM=INTERCEPT | SUBJECT(id) COVTYPE(VC).
+
+
+DATASET CLOSE workhours.
+
+
 GET FILE='C:\Users\11187439\Desktop\SPSS\C22_E3.sav'. 
 DATASET NAME C22 WINDOW=FRONT.
 
-MIXED guilt BY anchor vignette verdictsyst WITH prior
+MIXED guilt BY anchor vignette verdictsyst WITH pjaq
   /CRITERIA=DFMETHOD(SATTERTHWAITE)
-  /FIXED=anchor vignette verdictsyst anchor*vignette anchor*verdictsyst vignette*verdictsyst anchor*vignette*verdictsyst prior | SSTYPE(3) 
+  /FIXED=anchor vignette verdictsyst anchor*vignette anchor*verdictsyst vignette*verdictsyst anchor*vignette*verdictsyst pjaq | SSTYPE(3) 
   /METHOD=REML 
   /PRINT=SOLUTION 
   /RANDOM=INTERCEPT | SUBJECT(id) COVTYPE(VC).
@@ -23,4 +37,4 @@ MIXED y BY choc background
   /RANDOM=INTERCEPT choc | SUBJECT(rater*background) COVTYPE(VC).
 
 
-DATASET CLOSE chocolate.
+DATASET CLOSE chocolate.

docs/evaluations/08-problem-set.html

@@ -313,7 +313,7 @@ <h1 class="title display-7">Problem set 8</h1>
 <li>proper randomization based only on <code>Prior</code> scores</li>
 <li>equality of slopes</li>
 </ol></li>
-<li>If the variance are unequal, fit the model with unequal variances per group (see the <a href="example/ancova.html">ANCOVA example</a>) — otherwise proceed as usual.
+<li>If the variance are unequal, fit the model with unequal variances per group (see the <a href="../example/ancova.html">ANCOVA example</a>) — otherwise proceed as usual.
 <ul>
 <li>Look at the pairwise difference between between Boost+ and consensus only condition and report the results of this test.</li>
 <li>Does the conclusion of the test change relative to that reported in the paper?</li>

docs/evaluations/09-problem-set.html

@@ -318,7 +318,6 @@ <h2 id="toc-title">On this page</h2>
 
   <ul>
   <li><a href="#task-1" id="toc-task-1" class="nav-link active" data-scroll-target="#task-1">Task 1</a></li>
-  <li><a href="#task-2" id="toc-task-2" class="nav-link" data-scroll-target="#task-2">Task 2</a></li>
   </ul>
 </nav>
     </div>
@@ -336,33 +335,32 @@ <h1 class="title display-7">Problem set 9</h1>
 <li>a PDF report</li>
 <li>your code</li>
 </ul>
-<p>Look at the <a href="../example/effectsizepower.html">effect size and power</a>; this should be helpful in completing the problem set.</p>
+<p>Look at the <a href="../example/effectsizepower.html">effect size and power examples</a>; this should be helpful in completing the problem set.</p>
 <section id="task-1" class="level2">
 <h2 class="anchored" data-anchor-id="task-1">Task 1</h2>
 <p>The following text is a quote from <a href="https://osf.io/3467b">Investigating variation in replicability: A “Many Labs” Replication Project by Richard A. Klein, Kate A. Ratliff, and Brian A. Nosek (pp.&nbsp;13-14)</a> about a replication of <span class="citation" data-cites="Oppenheimer.Monin:2009">Oppenheimer &amp; Monin (<a href="#ref-Oppenheimer.Monin:2009" role="doc-biblioref">2009</a>)</span> work on the retrospective gambler fallacy.</p>
 <blockquote class="blockquote">
 <p>The differences between groups was reliable, omnibus <span class="math inline">\(F(2, 77) = 4.8\)</span>, […] Cohen’s <span class="math inline">\(f = 0.18\)</span>. Pairwise comparisons showed that all differences between conditions were reliable as well, <span class="math inline">\(t(47, 48, 57) = 1.94, 2.32, 2.65\)</span>; <span class="math inline">\(p &lt; .05\)</span>, Cohen’s <span class="math inline">\(d\)</span> = <span class="math inline">\((.56, .67, .69)\)</span>.</p>
 </blockquote>
 <ol type="1">
-<li>Using the information from the quote, compute the effect size for Cohen’s <span class="math inline">\(f\)</span> and <span class="math inline">\(\widehat{\omega}^2\)</span>.<a href="#fn1" class="footnote-ref" id="fnref1" role="doc-noteref"><sup>1</sup></a> Note that you won’t necessarily get the same value of Cohen’s <span class="math inline">\(f\)</span> that is reported.</li>
-<li>Compute the overall sample size necessary to replicate this study with a power of at least <span class="math inline">\(0.99:\)</span> compute the power for both the overall effect and the three pairwise <span class="math inline">\(t\)</span>-tests based on the reported values of Cohen’s <span class="math inline">\(d\)</span> and pick the maximum sample size needed for balanced design (i.e., gathering the same number of participants in each subgroup).</li>
+<li>Using the information from the quote, compute the effect size for Cohen’s <span class="math inline">\(f\)</span> and <span class="math inline">\(\widehat{\omega}^2\)</span>.<a href="#fn1" class="footnote-ref" id="fnref1" role="doc-noteref"><sup>1</sup></a> Note that you won’t normally get the same value of Cohen’s <span class="math inline">\(f\)</span> that is reported.</li>
+<li>Compute the overall sample size necessary to replicate this study with a power of at least <span class="math inline">\(0.99:\)</span> compute the sample sizes for both the overall effect (using the value you calculated in 1.) and the three pairwise <span class="math inline">\(t\)</span>-tests based on the reported values of Cohen’s <span class="math inline">\(d\)</span> and pick the maximum sample size needed for a balanced design, i.e., gathering the same number of participants in each subgroup.</li>
 <li>Read the results of the replication in <a href="https://psycnet.apa.org/fulltext/2014-20922-002.pdf">ManyLabs1</a>; these are illustrated in Figure 1 and summary statistics about standardized mean differences (Cohen’s <span class="math inline">\(d\)</span>) are reported in Table 2 <span class="citation" data-cites="ManyLab1:2014">(<a href="#ref-ManyLab1:2014" role="doc-biblioref">Klein et al., 2014</a>)</span> under retrospective gambler fallacy. Comment about the success of the replication.</li>
 <li>Based on observation of Figure 1 of <span class="citation" data-cites="ManyLab1:2014">Klein et al. (<a href="#ref-ManyLab1:2014" role="doc-biblioref">2014</a>)</span>, why might one object to using estimated effect size reported in peer-reviewed papers?<a href="#fn2" class="footnote-ref" id="fnref2" role="doc-noteref"><sup>2</sup></a></li>
 </ol>
-</section>
-<section id="task-2" class="level2">
-<h2 class="anchored" data-anchor-id="task-2">Task 2</h2>
-<p>Read the Section “Going Beyond Null Hypothesis Significance Testing” of the Academy of Management Journal 2024 <a href="https://journals.aom.org/doi/epub/10.5465/amj.2024.4004">editorial</a> <span class="citation" data-cites="Bliese:2024">(<a href="#ref-Bliese:2024" role="doc-biblioref">Bliese et al., 2024</a>)</span> in preparation for a in-class discussion.</p>
+<!--
+## Task 2
+
+Read the Section "Going Beyond Null Hypothesis Significance Testing" of the Academy of Management Journal 2024 [editorial](https://journals.aom.org/doi/epub/10.5465/amj.2024.4004) [@Bliese:2024] in preparation for a in-class discussion.
+
+-->
 
 
 
 </section>
 
 
 <div id="quarto-appendix" class="default"><section class="quarto-appendix-contents" role="doc-bibliography" id="quarto-bibliography"><h2 class="anchored quarto-appendix-heading">References</h2><div id="refs" class="references csl-bib-body hanging-indent" data-entry-spacing="0" data-line-spacing="2" role="list">
-<div id="ref-Bliese:2024" class="csl-entry" role="listitem">
-Bliese, P. D., Certo, S. T., Smith, A. D., Wang, M., &amp; Gruber, M. (2024). Strengthening theory–methods–data links. <em>Academy of Management Journal</em>, <em>67</em>(4), 893–902. <a href="https://doi.org/10.5465/amj.2024.4004">https://doi.org/10.5465/amj.2024.4004</a>
-</div>
 <div id="ref-ManyLab1:2014" class="csl-entry" role="listitem">
 Klein, R. A., Ratliff, K. A., Vianello, M., Adams Jr., R. B., Bahník, Š., Bernstein, M. J., Bocian, K., Brandt, M. J., Brooks, B., Brumbaugh, C. C., Cemalcilar, Z., Chandler, J., Cheong, W., Davis, W. E., Devos, T., Eisner, M., Frankowska, N., Furrow, D., Galliani, E. M., … Nosek, B. A. (2014). Investigating variation in replicability: A <span>“many labs”</span> replication project. <em>Social Psychology</em>, <em>45</em>, 142–152. <a href="https://doi.org/10.1027/1864-9335/a000178">https://doi.org/10.1027/1864-9335/a000178</a>
 </div>

docs/search.json

@@ -1208,7 +1208,7 @@
     "href": "evaluations/09-problem-set.html",
     "title": "Problem set 9",
     "section": "",
-    "text": "Submission information: please submit on ZoneCours\nLook at the effect size and power; this should be helpful in completing the problem set.",
+    "text": "Submission information: please submit on ZoneCours\nLook at the effect size and power examples; this should be helpful in completing the problem set.",
     "crumbs": [
       "Evaluations",
       "Problem sets",
@@ -1220,19 +1220,7 @@
     "href": "evaluations/09-problem-set.html#task-1",
     "title": "Problem set 9",
     "section": "Task 1",
-    "text": "Task 1\nThe following text is a quote from Investigating variation in replicability: A “Many Labs” Replication Project by Richard A. Klein, Kate A. Ratliff, and Brian A. Nosek (pp. 13-14) about a replication of Oppenheimer & Monin (2009) work on the retrospective gambler fallacy.\n\nThe differences between groups was reliable, omnibus \\(F(2, 77) = 4.8\\), […] Cohen’s \\(f = 0.18\\). Pairwise comparisons showed that all differences between conditions were reliable as well, \\(t(47, 48, 57) = 1.94, 2.32, 2.65\\); \\(p < .05\\), Cohen’s \\(d\\) = \\((.56, .67, .69)\\).\n\n\nUsing the information from the quote, compute the effect size for Cohen’s \\(f\\) and \\(\\widehat{\\omega}^2\\).1 Note that you won’t necessarily get the same value of Cohen’s \\(f\\) that is reported.\nCompute the overall sample size necessary to replicate this study with a power of at least \\(0.99:\\) compute the power for both the overall effect and the three pairwise \\(t\\)-tests based on the reported values of Cohen’s \\(d\\) and pick the maximum sample size needed for balanced design (i.e., gathering the same number of participants in each subgroup).\nRead the results of the replication in ManyLabs1; these are illustrated in Figure 1 and summary statistics about standardized mean differences (Cohen’s \\(d\\)) are reported in Table 2 (Klein et al., 2014) under retrospective gambler fallacy. Comment about the success of the replication.\nBased on observation of Figure 1 of Klein et al. (2014), why might one object to using estimated effect size reported in peer-reviewed papers?2",
-    "crumbs": [
-      "Evaluations",
-      "Problem sets",
-      "Problem set 9"
-    ]
-  },
-  {
-    "objectID": "evaluations/09-problem-set.html#task-2",
-    "href": "evaluations/09-problem-set.html#task-2",
-    "title": "Problem set 9",
-    "section": "Task 2",
-    "text": "Task 2\nRead the Section “Going Beyond Null Hypothesis Significance Testing” of the Academy of Management Journal 2024 editorial (Bliese et al., 2024) in preparation for a in-class discussion.",
+    "text": "Task 1\nThe following text is a quote from Investigating variation in replicability: A “Many Labs” Replication Project by Richard A. Klein, Kate A. Ratliff, and Brian A. Nosek (pp. 13-14) about a replication of Oppenheimer & Monin (2009) work on the retrospective gambler fallacy.\n\nThe differences between groups was reliable, omnibus \\(F(2, 77) = 4.8\\), […] Cohen’s \\(f = 0.18\\). Pairwise comparisons showed that all differences between conditions were reliable as well, \\(t(47, 48, 57) = 1.94, 2.32, 2.65\\); \\(p < .05\\), Cohen’s \\(d\\) = \\((.56, .67, .69)\\).\n\n\nUsing the information from the quote, compute the effect size for Cohen’s \\(f\\) and \\(\\widehat{\\omega}^2\\).1 Note that you won’t normally get the same value of Cohen’s \\(f\\) that is reported.\nCompute the overall sample size necessary to replicate this study with a power of at least \\(0.99:\\) compute the sample sizes for both the overall effect (using the value you calculated in 1.) and the three pairwise \\(t\\)-tests based on the reported values of Cohen’s \\(d\\) and pick the maximum sample size needed for a balanced design, i.e., gathering the same number of participants in each subgroup.\nRead the results of the replication in ManyLabs1; these are illustrated in Figure 1 and summary statistics about standardized mean differences (Cohen’s \\(d\\)) are reported in Table 2 (Klein et al., 2014) under retrospective gambler fallacy. Comment about the success of the replication.\nBased on observation of Figure 1 of Klein et al. (2014), why might one object to using estimated effect size reported in peer-reviewed papers?2",
     "crumbs": [
       "Evaluations",
       "Problem sets",

docs/sitemap.xml

@@ -86,7 +86,7 @@
   </url>
   <url>
     <loc>https://lbelzile.github.io/experimental/evaluations/09-problem-set.html</loc>
-    <lastmod>2024-08-25T13:41:55.206Z</lastmod>
+    <lastmod>2024-11-20T16:35:37.605Z</lastmod>
   </url>
   <url>
     <loc>https://lbelzile.github.io/experimental/evaluations/07-problem-set.html</loc>
@@ -114,7 +114,7 @@
   </url>
   <url>
     <loc>https://lbelzile.github.io/experimental/evaluations/08-problem-set.html</loc>
-    <lastmod>2024-07-03T21:13:17.214Z</lastmod>
+    <lastmod>2024-11-20T16:35:37.605Z</lastmod>
   </url>
   <url>
     <loc>https://lbelzile.github.io/experimental/evaluations/final-exam.html</loc>

docs/slides/10-slides.html

@@ -4,7 +4,7 @@
     <title>Introduction to mixed models</title>
     <meta charset="utf-8" />
     <meta name="author" content="Léo Belzile" />
-    <meta name="date" content="2024-10-31" />
+    <meta name="date" content="2024-11-08" />
     <script src="libs/header-attrs-2.28/header-attrs.js"></script>
     <link href="libs/remark-css-0.0.1/default.css" rel="stylesheet" />
     <link href="libs/tile-view-0.2.6/tile-view.css" rel="stylesheet" />
@@ -593,23 +593,6 @@
 
 ---
 
-# Formulae in **R**
-
-**R** uses the following notation 
-
-- `group1/group2` means `group2` is nested within `group1`. 
-
-The formula expands to `group1 + group1:group2`.
-
-- `group1*group2` means `group` and `group2` are **crossed**
-
-The formula is a shorthand for `group1 + group2 + group1:group2`.
-
-.small[
-To fit the model, identifiers of subjects must be declared as factors (categorical variables).
-]
-
----
 
 # Specifying interactions
 
@@ -648,7 +631,7 @@
 anova(ddf = "Kenward-Roger", # other option is "Satterthwaite"
   lmerTest::lmer(
     data = AA21 |&gt; dplyr::filter(latency &gt; -40), 
-    latency ~ stimulus + (1|id) + (1|id:stimulus)))
+    latency ~ stimulus + (1 | id) + (1 | id:stimulus)))
 ```
 
 ```
@@ -671,6 +654,15 @@
 &lt;img src="img/10/Hasse_diagram.png" width="60%" style="display: block; margin: auto;" /&gt;
 
 
+---
+
+# Example: chocolate rating
+
+Example from L. Meier, adapted from Oehlert (2010)
+
+&gt; A group of 10 rural and 10 urban raters rated 4 different chocolate types. Every rater got to eat two samples from the same chocolate type in random order.
+
+
 ---
 
 # Example: Curley et al. (2022)
@@ -683,14 +675,7 @@
 
 ]
 
----
-
-# Example: chocolate rating
-
-Example from L. Meier, adapted from Oehlert (2010)
-
-&gt; A group of 10 rural and 10 urban raters rated 4 different chocolate types. Every rater got to eat two samples from the same chocolate type in random order.
-
+This study is an instance of **incomplete** design.
     </textarea>
 <style data-target="print-only">@media screen {.remark-slide-container{display:block;}.remark-slide-scaler{box-shadow:none;}}</style>
 <script src="libs/remark-latest.min.js"></script>

evaluations/08-problem-set.qmd

@@ -20,7 +20,7 @@ We consider Experiment 3 of @vanStekelenburg:2021; ([download the paper](https:/
    a. equal variance per experimental condition with the ANCOVA
    b. proper randomization based only on `Prior` scores
    c. equality of slopes
-6. If the variance are unequal, fit the model with unequal variances per group (see the [ANCOVA example](example/ancova.html)) --- otherwise proceed as usual.
+6. If the variance are unequal, fit the model with unequal variances per group (see the [ANCOVA example](/example/ancova.html)) --- otherwise proceed as usual.
    - Look at the pairwise difference between between Boost+ and consensus only condition and report the results of this test.
    - Does the conclusion of the test change relative to that reported in the paper?