5.8 Warning

Users should also note that sometimes our GUI shuts down. In our experience, this is due to computational issues arising from the implicit commands we call when estimating certain models. These issues may include computationally singular systems, missing values where TRUE/FALSE are needed, L-BFGS-B requiring finite values for “fn”, NA/NaN/Inf values, or errors in backsolve. These issues can sometimes be resolved by adjusting the dataset, such as avoiding high levels of multicollinearity.

It should also be noted that when warning messages are displayed in our GUI, there is a high likelihood of convergence issues with the posterior chains. Therefore, the results may not be trustworthy. Users can identify these problems by checking the console in their RStudio sections, where the specific folder/file where the issue occurred will be specified. In any case, we would appreciate your feedback to improve and enhance our GUI.

We should also mention that there are many ways to improve the codes presented in this book, and particularly, the following five chapters. For instance, the MCMCpack and bayesm packages perform most of the matrix operations in C++ using the Rcpp package. This substantially speeds up the algorithms compared to the codes presented in the next chapters when we program from scratch the samplers. We could further improve the computational times of our codes using parallel computing and the Rcpp package, but this requires more advanced skills that are not covered in this book.