Questions About igt_vpp Function in hBayesDM Package

[Mingcong Tang]

Hello,

I'm working with the igt_vpp function to model data from an Iowa Gambling Task (IGT) study. While running the model, I noticed that the outputs appear to be identical regardless of whether the inc_postpred argument is set to TRUE or FALSE. Specifically, I expected differences in the output based on whether posterior predictive checks were included, but this does not seem to be the case. I was wondering if it could be a bug.

Here is an example of the code I used:

output1 <- igt_vpp(
  data = my_data, niter = 4000, nwarmup = 2000, nchain = 4, ncore = 4,
  nthin = 1,
  inits = "random",
  indPars = "mean",
  modelRegressor = FALSE,
  vb = FALSE,
  inc_postpred = FALSE,
  adapt_delta = 0.95,
  stepsize = 1,
  max_treedepth = 10)

output2 <- igt_vpp(
  data = my_data, niter = 4000, nwarmup = 2000, nchain = 4, ncore = 4,
  nthin = 1,
  inits = "random",
  indPars = "mean",
  modelRegressor = FALSE,
  vb = FALSE,
  inc_postpred = TRUE,
  adapt_delta = 0.95,
  stepsize = 1,
  max_treedepth = 10)

Thank you very much for your time and clarification!

Best regards,
Mingcong

[wooyou...@gmail.com]

Hi Mingcong, 

It's true that posterior distributions are not affected by the argument. Please see the link below for more details about the "inc_postpred" argument. It's just for deciding whether to save the posterior predictive checks or not. 

https://ccs-lab.github.io/hBayesDM/articles/getting_started.html#posterior-predictive-checks

Best,
Young