Merge branch 'develop' of https://github.com/stan-dev/stan into bugfi…

…x/issue-530-no-model-parameters

make/tests

@@ -194,6 +194,7 @@ $(patsubst %.stan,%.cpp,$(filter src/test/test-models/no-main/%,$(TEST_MODELS)))
 ##
 test/unit/io/mcmc_writer.o: src/test/test-models/no-main/io_example.cpp
 test/unit/gm/command_init.o: src/test/test-models/no-main/gm/test_lp.cpp
+test/unit/gm/command_write_iteration.o: src/test/test-models/no-main/gm/test_lp.cpp
 test/unit/gm/generator.o: src/test/test-models/no-main/gm/test_lp.cpp
 
 test/unit/model/util.o: src/test/test-models/no-main/model/valid.cpp src/test/test-models/no-main/model/domain_fail.cpp
@@ -216,6 +217,7 @@ test/CmdStan/command.o: \
                $(addsuffix $(EXE), printer domain_fail value_fail proper))
 test/CmdStan/gm/argument_configuration.o: src/test/test-models/compiled/CmdStan/test_model$(EXE)
 test/CmdStan/fixed_param_sampler.o: src/test/test-models/compiled/CmdStan/empty$(EXE) src/test/test-models/compiled/CmdStan/proper$(EXE)
+test/CmdStan/optimization_output.o: src/test/test-models/compiled/CmdStan/optimization_output$(EXE)
 
 ##
 # CmdStan specific targets. Should be removed in the future.

src/docs/bibtex/all.bib

@@ -288,13 +288,14 @@ @incollection{Gay:2005
 }
 
 
-@book{GelmanCarlinSternRubin:2003,
+@book{GelmanEtAl:2013, 
 	Address = {London},
-	Author = {Andrew Gelman and J.~B.~Carlin and Hal S.~Stern and Donald B.~Rubin},
-	Edition = {2nd},
-	Publisher = {CRC Press},
+	Author = {Andrew Gelman and J.~B.~Carlin and Hal S.~Stern and
+	David B.~Dunson and Aki Vehtari and Donald B.~Rubin},
+	Edition = {Third},
+	Publisher = {Chapman \&Hall/CRC Press},
 	Title = {Bayesian Data Analysis},
-	Year = {2003}}
+	Year = {2013}}
 
 @book{GelmanHill:2007,
 	Address = {Cambridge, United Kingdom},
@@ -351,6 +352,16 @@ @article{Hastings:1970
     Pages = {97--109}
 }
 
+@article{HoerlKennard:1970,
+  author = {Arthur E.\ Hoerl and Robert W.\ Kennard},
+  year = {1970},
+  title = {Ridge regression: biased estimation for nonorthogonal problems},
+  journal = {Technometrics},  
+  volume = {12},
+  number = {1},  
+  pages = {55--67}
+}
+
 @article{Hoffman-Gelman:2011,
     Author = {Matthew D. Hoffman and Andrew Gelman},
     Title = {The No-{U}-Turn Sampler: Adaptively Setting Path 
@@ -363,7 +374,7 @@ @article{Hoffman-Gelman:2011
     Year = {2011}
 }
 
-@article{Hoffman-Gelman:2013,
+@article{Hoffman-Gelman:2014,
     Author = {Matthew D. Hoffman and Andrew Gelman},
     Title = {The No-{U}-Turn Sampler: Adaptively Setting Path 
              Lengths in {H}amiltonian {M}onte {C}arlo},
@@ -387,6 +398,19 @@ @book{HuntThomas:99
   publisher = {Addison-Wesley}
 }
 
+@inproceedings{JamesStein:1961,
+  author = {W.\ James and Charles Stein},
+  year = {1961},
+  title = {Estimation with quadratic loss}, 
+  editor = {Jerzey Neyman},
+  booktitle = {Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability},
+  volume = {1},
+  pages = {361--379},
+  publisher = {University of California Press}
+}
+
+
+
 @article{Lambert:1992,  
   author = {Diane Lambert},  
   year = {1992},
@@ -462,6 +486,17 @@ @article{Neal:1994
    Pages = {194--203}
 }
 
+@article{Neal:1996b,
+  author = {Radford M. Neal}, 
+  year = {1996},
+  title = {Sampling from multimodal distributions using tempered
+           transitions},
+  journal = {Statistics and Computing},
+  volume = {6},
+  number = {4},
+  pages = {353--366}
+}
+
 @techreport{Neal:1997,
   author = {Neal, Radford M.},
   year = {1997},
@@ -617,6 +652,25 @@ @article{SmithSpiegelhalterThomas:1995
   pages = {2685--2699}
 }
 
+@article{SwendsenWang:1986,  
+  author = {Robert H. Swendsen and Jian-Sheng Wang},  
+  year = {1986},  
+  title = {Replica Monte Carlo simulation of spin glasses},
+  journal = {Physical Review Letters},
+  volume = {57},
+  pages = {2607--2609}
+}
+
+@article{Tibshirani:1996,
+  author = {Robert Tibshirani},
+  year = {1996},  
+  title = {Regression shrinkage and selection via the lasso},
+  journal = {Journal of the Royal Statistical Society, Series B},
+  volume = {58},
+  number = {1},
+  pages = {267--288}
+}
+
 @techreport{TokudaGoodrichVanMechelen:2010,
 	Author = {Tomoki Tokuda and Ben Goodrich and Iven Van~Mechelen and Andrew Gelman and Francis Tuerlinckx},
 	Date-Added = {2011-10-11 09:21:36 -0400},
@@ -665,6 +719,15 @@ @inproceedings{Zhang:2011
   year={2011}
 }
 
+@article{ZouHastie:2005,
+  author = {Hui Zou and Trevor Hastie}, 
+  year = {2005},
+  title = {Regularization and variable selection via the elastic net},
+  journal = {Journal of the Royal Statistical Society, Series B},
+  volume = {67},  
+  number = {2},
+  pages = {301--320}
+}
 
 %%%% ==================== OLDER BIB ITEMS ======================
 
@@ -1999,21 +2062,6 @@ @book{Dye:1966
 	Title = {Politics, Economics, and the Public: Political Outcomes in the American States},
 	Year = {1966}}
 
-@article{EfronMorris:1975,
-	Author = {Efron, B. and Morris, C.},
-	Journal = {Journal of the American Statistical Association},
-	Pages = {311--319},
-	Title = {Data analysis using Stein's estimator and its generalizations},
-	Volume = {70},
-	Year = {1975}}
-
-@book{EfronTibshirani:1993,
-	Address = {London},
-	Author = {Efron, B. and Tibshirani, R. J.},
-	Publisher = {Chapman and Hall},
-	Title = {An Introduction to the Bootstrap},
-	Year = {1993}}
-
 @book{Ehrenberg:1978,
 	Address = {New York},
 	Author = {Ehrenberg, A. S. C.},
@@ -3220,11 +3268,24 @@ @book{HastieTibshirani:1990
 	Year = {1990}}
 
 @book{HastieTibshiraniFriedman:2002,
-	Address = {New York},
 	Author = {Hastie, T. J. and Tibshirani, R. J. and Friedman, J.},
+	Year = {2002},
+	Title = {The Elements of Statistical Learning: Data Mining,
+                 Inference, and Prediction},
+        Edition = {First},
 	Publisher = {Springer-Verlag},
-	Title = {The Elements of Statistical Learning: Data Mining, Inference, and Prediction},
-	Year = {2002}}
+	Address = {New York}
+}
+
+@book{HastieTibshiraniFriedman:2009,
+	Author = {Trevor Hastie and Robert Tibshirani and Jerome Friedman},
+	Year = {2009},
+	Title = {The Elements of Statistical Learning: Data Mining,
+                 Inference, and Prediction},
+        Edition = {Second},
+	Publisher = {Springer-Verlag},
+	Address = {New York}
+}
 
 @article{HauerNgLovell:1988,
 	Author = {Hauer, E. and Ng, J. C. N. and Lovell, J.},
@@ -7336,6 +7397,7 @@ @article{EfronMorris:1975
 	Volume = {70},
 	Year = {1975}}
 
+
 @book{EfronTibshirani:1993,
 	Address = {London},
 	Author = {Efron, B. and Tibshirani, R. J.},

src/docs/stan-models/build.xml

@@ -3,7 +3,7 @@
 <project name="Stan Model PDF" default="pdf" basedir=".">
 
   <property name="version"  
-            value="2.1.0"/>
+            value="2.2.0"/>
 
   <path id="classpath">
     <pathelement location="stan-models.jar"/>

src/docs/stan-models/models.sty

@@ -56,5 +56,5 @@
 
 \sloppy % no line overruns
 
-\title{{\Huge\bf Stan Example Models} \\[12pt] Version 2.1.0 \\[48pt]}
+\title{{\Huge\bf Stan Example Models} \\[12pt] Version 2.2.0 \\[48pt]}
 \author{Stan Development Team \vfill \today}

src/docs/stan-models/src/MakeModelTex.java

@@ -56,7 +56,7 @@ static void writeStart(Writer texWriter)
         texWriter.write("\\\\[2in]" + EOL);
         texWriter.write("{\\Huge\\bf Stan Example Models}" + EOL);
         texWriter.write("\\\\[18pt]" + EOL);
-        texWriter.write("{\\large Version 2.1.0}" + EOL);
+        texWriter.write("{\\large Version 2.2.0}" + EOL);
         texWriter.write("\\\\[12pt]" + EOL);
         texWriter.write("{\\large \\today}" + EOL);
         texWriter.write("\\\\[1in]" + EOL);

src/docs/stan-reference/R/non-identified-plot.R

@@ -0,0 +1,137 @@
+library(ggplot2);
+library(grid);
+
+N <- 100;
+y <- rnorm(N);
+
+loc_sum <- function(lambda1,lambda2) {
+  sum(dnorm(y,lambda1 + lambda2, 1, log=TRUE));
+}
+
+loc_sum_prior <- function(lambda1,lambda2) {
+  sum(dnorm(y,lambda1 + lambda2, 1, log=TRUE)) + 
+  sum(dnorm(c(lambda1,lambda2),0,1,log=TRUE));
+}
+
+one_param <- function(xs) {  
+  dta <- rnorm(100,0,1);
+  result <- rep(NA, length(xs));
+  i <- 1;
+  for (x in xs) {
+    result[i] <- sum(dnorm(dta,xs[i],1,log=TRUE));
+    i <- i + 1;
+  } 
+  return(result);
+}
+p_one_param <-
+  ggplot(data.frame(mu=c(-20, 20)), aes(mu)) +
+  labs(title = "Proper Posterior (without Prior)\n") + 
+  stat_function(fun=one_param) +
+  labs(x=expression(mu), y="log p") +
+  theme(aspect.ratio=1,
+        panel.border=element_blank(),
+        plot.margin=unit(c(0,0,0,0),"cm"),
+        text=element_text(size=28),
+        axis.title=element_text(size=32));
+png(res=100,height=800,width=900,file="one_param_identified.png");
+print(p_one_param);
+dev.off();
+
+
+
+
+
+K <- 500;
+ub <- 25;
+lambda_1 <- seq(-ub,ub,len=K);
+lambda_2 <- seq(-ub,ub,len=K);
+
+v_lambda_1 <- rep(NA,K^2);
+v_lambda_2 <- rep(NA,K^2);
+v_density <- rep(NA,K^2);
+
+# use prior (loc_sum_prior)
+pos <- 1;
+for (m in 1:K) {
+  for (n in 1:K) {
+    v_lambda_1[pos] <- lambda_1[m];
+    v_lambda_2[pos] <- lambda_2[n];
+    v_density[pos] <- loc_sum_prior(lambda_1[m], lambda_2[n]);
+    pos <- pos + 1;
+  }
+}
+df_vals <- list(lambda_1=v_lambda_1,lambda_2=v_lambda_2,log_density=v_density);
+df_prior <- as.data.frame(df_vals);
+
+# plot and save as png
+
+p_prior <- 
+ ggplot(df_prior,aes(x=lambda_1,y=lambda_2,fill=log_density)) +
+ labs(title = "Proper Posterior (with Prior)\n") + 
+ geom_tile() +
+ labs(x=expression(lambda[1]), y=expression(lambda[2])) + 
+ scale_fill_gradient2("log p",
+                      limits=c(-600,-120),
+                      midpoint=-400, low="gray95", high="darkblue",
+                      mid="lightyellow", na.value="transparent",      
+                      breaks=c(-200, -400, -600),   
+                      labels=c("-200", "-400", "-600")) +
+  scale_x_continuous(expand = c(0, 0)) +
+  scale_y_continuous(expand = c(0, 0)) +
+  theme(aspect.ratio=1,
+        panel.border=element_blank(),
+        plot.margin=unit(c(0,0,0,0),"cm"),
+        text=element_text(size=28),
+        axis.title=element_text(size=32),
+        legend.title=element_text(size=30,face="plain",color="gray25"), 
+        legend.text=element_text(size=20,color="gray40"),
+        legend.margin=unit(0.3,"cm") )
+
+png(res=100,height=800,width=900,file="non-identified-plus-prior.png");
+plot(p_prior)
+dev.off();
+plot(p_prior)
+
+# no prior (loc_sum)
+pos <- 1;
+for (m in 1:K) {
+  for (n in 1:K) {
+    v_density[pos] <- loc_sum(lambda_1[m], lambda_2[n]);
+    pos <- pos + 1;
+  }
+}
+df_vals <- list(lambda_1=v_lambda_1,lambda_2=v_lambda_2,log_density=v_density);
+df_no_prior <- as.data.frame(df_vals);
+
+# plot and save as png
+
+p_no_prior <- 
+ ggplot(df_no_prior,aes(x=lambda_1,y=lambda_2,fill=log_density)) +
+ labs(title = "Improper Posterior (without Prior)\n") + 
+ geom_tile() +
+ labs(x=expression(lambda[1]), y=expression(lambda[2])) +
+ scale_x_continuous(expand = c(0, 0)) +
+ scale_y_continuous(expand = c(0, 0)) +
+ scale_fill_gradient2("log p",
+                      limits=c(-600,-120),
+                      midpoint=-400, low="gray95", high="darkblue",
+                      mid="lightyellow", na.value="transparent",
+                      breaks=c(-200, -400, -600),   
+                      labels=c("-200", "-400", "-600") ) +
+ theme(aspect.ratio=1,
+        panel.border=element_blank(),
+        plot.margin=unit(c(0,0,0,0),"cm"),
+        text=element_text(size=28),
+        axis.title=element_text(size=32),
+        legend.title=element_text(size=30,face="plain",color="gray25"), 
+        legend.text=element_text(size=20,color="gray40"),
+        legend.margin=unit(0.3,"cm") )
+
+
+  theme(aspect.ratio=1,text=element_text(size=28),
+        axis.title=element_text(size=32),
+        legend.margin=unit(0.3,"cm"));
+
+png(res=100,height=800,width=900,file="non-identified.png")
+print(p_no_prior)
+dev.off();