Altering the estimation process

Bayesian Regression Modeling with rstanarm

Jake Thompson

Psychometrician, ATLAS, University of Kansas

Divergent transitions

1: There were 15 divergent transitions after warmup. Increasing adapt_delta
above 0.8 may help.
  • Too big of steps in the estimator
  • Adjust step size
stan_model <- stan_glm(popularity ~ song_age, data = songs,
  control = list(adapt_delta = 0.95))

stan_model <- stan_glm(popularity ~ song_age, data = songs,
  control = list(adapt_delta = 0.99))
Bayesian Regression Modeling with rstanarm

Exceeding the Maximum Treedepth

Chain 1 reached the maximum tree depth
  • Sample evaluates branches and looks for a good place to "U-Turn"
  • Max tree depth indicates poor efficiency
stan_model <- stan_glm(popularity ~ song_age, data = songs,
  control = list(max_treedepth = 10))

stan_model <- stan_glm(popularity ~ song_age, data = songs,
  control = list(max_treedepth = 15))
Bayesian Regression Modeling with rstanarm

Tuning the estimation

  • Estimation errors are threats to the validity of the model
  • Although complicated, these errors can be addressed easily
Bayesian Regression Modeling with rstanarm

Let's practice!

Bayesian Regression Modeling with rstanarm

Preparing Video For Download...