Summary:
I've noticed neg_binomial_2_lpmf sometimes returns values > 0. These would correspond to pmf > 1 (which is not possible). It seems to occur only when phi is a very large number (on the order of 1e15).
Description:
When neg_binomial_2_lpmf returns positive values, constraints are violated, samples are rejected. Sometimes very large phi occur due to random initialization or during sampling.
Reproducible Steps:
R script and Stan model for a reproducible example are attached.
negBin2_error.stan.txt
negBin2_error.r.txt
4 inits are given showing that large phi generates errors. This model is cut down from a more complex model so some choices may seem odd (like checking lpmf > 0 or calculating log prob of data in the transformed parameters section).
Current Output:
Reject statement is activated in chains initialized with large phi showing lmpf value calculated (and inputs).
Expected Output:
Never execute this reject with Y, mu, and phi all well-defined.
Additional Information:
R version 3.3.0 (2016-05-03)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1
locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252 LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] MASS_7.3-45 shinystan_2.2.1 shiny_0.13.2 plyr_1.8.3 rstan_2.12.1 StanHeaders_2.12.0 ggplot2_2.1.0
Current Version:
v2.12.0
Summary:
I've noticed neg_binomial_2_lpmf sometimes returns values > 0. These would correspond to pmf > 1 (which is not possible). It seems to occur only when phi is a very large number (on the order of 1e15).
Description:
When neg_binomial_2_lpmf returns positive values, constraints are violated, samples are rejected. Sometimes very large phi occur due to random initialization or during sampling.
Reproducible Steps:
R script and Stan model for a reproducible example are attached.
negBin2_error.stan.txt
negBin2_error.r.txt
4 inits are given showing that large phi generates errors. This model is cut down from a more complex model so some choices may seem odd (like checking lpmf > 0 or calculating log prob of data in the transformed parameters section).
Current Output:
Reject statement is activated in chains initialized with large phi showing lmpf value calculated (and inputs).
Expected Output:
Never execute this reject with Y, mu, and phi all well-defined.
Additional Information:
R version 3.3.0 (2016-05-03)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1
locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252 LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C LC_TIME=English_United States.1252
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] MASS_7.3-45 shinystan_2.2.1 shiny_0.13.2 plyr_1.8.3 rstan_2.12.1 StanHeaders_2.12.0 ggplot2_2.1.0
Current Version:
v2.12.0