Syncing in new data examples

DESCRIPTION

@@ -1,10 +1,10 @@
 Package: rethinking
 Type: Package
 Title: Statistical Rethinking book package
-Version: 1.10
+Version: 1.11
 Date: 2012-09-27
 Author: Richard McElreath
 Maintainer: Richard McElreath <[email protected]>
 Depends: MASS, bbmle, coda
 Description: Utilities for fitting and comparing models
-License: GPL version 2 or newer
+License: GPLv3

R/map.r

@@ -0,0 +1,37 @@
+# MAP - maximum a posteriori
+# uses Bolker's mle2 (bbmle) as the engine, just smuggling in priors as penalized likelihood
+
+# format for priors:
+# list( formula1 , formula2 , ... )
+# formulas are in format of the likelihood formula, i.e.
+# parameter ~ dfunction( pars )
+# e.g.:
+# a ~ dnorm(0,1)
+# s ~ dunif(0,100)
+
+# map() then builds the likelihood formula and prior formulas into a unified enclosure that returns penalized log-likelihood. optim() does the rest.
+
+map <- function( formula , data , start , prior , ... ) {
+
+    require(bbmle)
+
+    # check parameters
+
+    if ( missing( formula ) | missing( data ) | missing( start ) ) {
+        stop( "Must specify formula, data, and start." )
+    }
+    if ( missing( prior ) ) {
+        # okay, but mark as NULL
+        prior <- NULL
+    }
+
+    # check that none of the declared priors are for a parameter not in start list
+
+    #################
+    # build enclosure
+
+}
+
+# testing examples
+# data(cars)
+# fit <- map( dist ~ dnorm( mean=a+b*speed , sd=sigma ) , data=cars , start=list( a=mean(cars$dist) , b=0 , sigma=sd(cars$dist) ) , prior=list( b ~ dnorm(0,1) , sigma ~ dunif(0,100) ) )