minor update + PIT ecdf

demos_ch6/demo6_2.R

@@ -50,7 +50,6 @@ ggplot(data = Tyr) +
                  color = 'black', binwidth = 1) +
   geom_vline(aes(xintercept = x), data = data.frame(x = Ty),
              color = 'red') +
-  labs(x = TeX('Number of changes in \\mathit{y} and \\mathit{y}^{\\mathrm{rep}}'),
+  labs(x = TeX(r"(Number of changes in $\textit{y}$ and  $\textit{y}^{ \textrm{rep}}$)"),
        y = '', title = title1) +
   scale_y_continuous(breaks=NULL)
-

demos_ch6/demo6_2.Rmd

@@ -69,7 +69,7 @@ ggplot(data = Tyr) +
                  color = 'black', binwidth = 1) +
   geom_vline(aes(xintercept = x), data = data.frame(x = Ty),
              color = 'red') +
-  labs(x = TeX('Number of changes in \\mathit{y} and \\mathit{y}^{\\mathrm{rep}}'),
+  labs(x = TeX(r"(Number of changes in $\textit{y}$ and  $\textit{y}^{ \textrm{rep}}$)"),
        y = '', title = title1) +
   scale_y_continuous(breaks=NULL)
 ```

demos_ch6/demo6_4.R

@@ -19,7 +19,8 @@
 #' ggplot2 is used for plotting, tidyr for manipulating data frames
 #+ setup, message=FALSE, error=FALSE, warning=FALSE
 library(ggplot2)
-theme_set(theme_minimal())
+library(bayesplot)
+theme_set(bayesplot::theme_default(base_family = "sans"))
 library(tidyr)
 library(latex2exp)
 library(rprojroot)
@@ -38,15 +39,23 @@ Ty <- data.frame(x = pt((y - my)/(sqrt(1+1/n)*s), n-1))
 
 #* Plot histogram of PIT values. Ideally histogram should be close to uniform.
 title1 <- 'Light speed example
-distribution of marginal posterior tail-values'
+distribution of predictive distribution tail-values'
+
 ggplot(data = Ty) +
   geom_histogram(aes(x = x), fill = 'steelblue',
                  color = 'black', binwidth = 0.05) +
   coord_cartesian(xlim = c(0, 1)) +
-  labs(x = TeX('\\mathit{p}(\\mathit{y}^{\\mathrm{rep}}_{\\mathit{i}} < \\mathit{y_i} | \\mathit{y})'),
+  labs(x = TeX(r"($\textit{p}(\textit{y}^{\textrm{rep}}_{\textit{i}} < \textit{y_i} | \textit{y})$)"),
        y = '', title = title1) +
   scale_y_continuous(breaks=NULL)
 
+#* Plot ECDF of PIT values. Ideally ECDF should be close to diagonal line
+ggplot(data=data.frame(x=Ty), aes(x)) +
+  stat_ecdf(geom = "step", color=4) +
+  xlim(c(0,1))+
+  labs(x="Observed PIT values", y="ECDF")+
+  annotate(geom="segment",x=0,y=0,xend=1,yend=1)
+
 #' Repeat the PIT checking after removing two "outliers"
 y <- y[y>0]
 #' Sufficient statistics
@@ -60,11 +69,18 @@ Ty <- data.frame(x = pt((y - my)/(sqrt(1+1/n)*s), n-1))
 
 #' Plot histogram of PIT values. Ideally histogram should be close to uniform.
 title1 <- 'Light speed example
-distribution of marginal posterior tail-values'
+distribution of predictive distribution tail-values'
 ggplot(data = Ty) +
   geom_histogram(aes(x = x), fill = 'steelblue',
                  color = 'black', binwidth = 0.05) +
   coord_cartesian(xlim = c(0, 1)) +
-  labs(x = TeX('\\mathit{p}(\\mathit{y}^{\\mathrm{rep}}_{\\mathit{i}} < \\mathit{y_i} | \\mathit{y})'),
+  labs(x = TeX(r"($\textit{p}(\textit{y}^{\textrm{rep}}_{\textit{i}} < \textit{y_i} | \textit{y})$)"),
        y = '', title = title1) +
   scale_y_continuous(breaks=NULL)
+
+#* Plot ECDF of PIT values. Ideally ECDF should be close to diagonal line
+ggplot(data=data.frame(x=Ty), aes(x)) +
+  stat_ecdf(geom = "step", color=4) +
+  xlim(c(0,1))+
+  labs(x="Observed PIT values", y="ECDF")+
+  annotate(geom="segment",x=0,y=0,xend=1,yend=1)

demos_ch6/demo6_4.Rmd

@@ -18,7 +18,8 @@ ggplot2 is used for plotting, tidyr for manipulating data frames
 
 ```{r setup, message=FALSE, error=FALSE, warning=FALSE}
 library(ggplot2)
-theme_set(theme_minimal())
+library(bayesplot)
+theme_set(bayesplot::theme_default(base_family = "sans"))
 library(tidyr)
 library(latex2exp)
 library(rprojroot)
@@ -47,14 +48,22 @@ Ty <- data.frame(x = pt((y - my)/(sqrt(1+1/n)*s), n-1))
 
 #* Plot histogram of PIT values. Ideally histogram should be close to uniform.
 title1 <- 'Light speed example
-distribution of marginal posterior tail-values'
+distribution of predictive distribution tail-values'
+
 ggplot(data = Ty) +
   geom_histogram(aes(x = x), fill = 'steelblue',
                  color = 'black', binwidth = 0.05) +
   coord_cartesian(xlim = c(0, 1)) +
-  labs(x = TeX('\\mathit{p}(\\mathit{y}^{\\mathrm{rep}}_{\\mathit{i}} < \\mathit{y_i} | \\mathit{y})'),
+  labs(x = TeX(r"($\textit{p}(\textit{y}^{\textrm{rep}}_{\textit{i}} < \textit{y_i} | \textit{y})$)"),
        y = '', title = title1) +
   scale_y_continuous(breaks=NULL)
+
+#* Plot ECDF of PIT values. Ideally ECDF should be close to diagonal line
+ggplot(data=data.frame(x=Ty), aes(x)) +
+  stat_ecdf(geom = "step", color=4) +
+  xlim(c(0,1))+
+  labs(x="Observed PIT values", y="ECDF")+
+  annotate(geom="segment",x=0,y=0,xend=1,yend=1)
 ```
 
 Repeat the PIT checking after removing two "outliers"
@@ -82,13 +91,20 @@ Plot histogram of PIT values. Ideally histogram should be close to uniform.
 
 ```{r }
 title1 <- 'Light speed example
-distribution of marginal posterior tail-values'
+distribution of predictive distribution tail-values'
 ggplot(data = Ty) +
   geom_histogram(aes(x = x), fill = 'steelblue',
                  color = 'black', binwidth = 0.05) +
   coord_cartesian(xlim = c(0, 1)) +
-  labs(x = TeX('\\mathit{p}(\\mathit{y}^{\\mathrm{rep}}_{\\mathit{i}} < \\mathit{y_i} | \\mathit{y})'),
+  labs(x = TeX(r"($\textit{p}(\textit{y}^{\textrm{rep}}_{\textit{i}} < \textit{y_i} | \textit{y})$)"),
        y = '', title = title1) +
   scale_y_continuous(breaks=NULL)
+
+#* Plot ECDF of PIT values. Ideally ECDF should be close to diagonal line
+ggplot(data=data.frame(x=Ty), aes(x)) +
+  stat_ecdf(geom = "step", color=4) +
+  xlim(c(0,1))+
+  labs(x="Observed PIT values", y="ECDF")+
+  annotate(geom="segment",x=0,y=0,xend=1,yend=1)
 ```