From the model, we first get the \(\bar{\theta}\) draws. A “trace plot” is produced for each of the 12 parameters in \(\bar{\theta}\). Ideally, the estimation should not show any serial correlation. For more explanations, Click Here (focus on the section about trace plot).

Next, I’ll plot the 12 parameters one by one.

setwd("C:/Users/chen/Desktop")
load("workspace.RData")

tb <- model$thetabardraw

trace_plot <- vector(mode="list",length=12)

library(ggpubr)
## Warning: package 'ggpubr' was built under R version 4.2.1
## Loading required package: ggplot2
for (i in 1:12) {
  trace_plot[[i]] <- plot(tb[i,1:10000], type = "l",
                          xlab = "Iterations", 
                          ylab = paste("Parameter","_",as.character(i),sep = ""))
}

From the plot, we can see that for Parameter 2 and 3, the convergence is not achieved. For other parameters, it seems the convergence starts only at 9,000 iterations.

It seems we simply do not have enough variation to estimate Parameter 2 and 3.