error in flash missing data imputation with ebnm_pn

stephenslab / ebnm

R package to fit Empirical Bayes Normal Means model.

https://stephenslab.github.io/ebnm/

12 stars 9 forks source link

Closed NKweiwang closed 6 years ago

NKweiwang commented 6 years ago

I tried flash for missing data imputation with ebnm_fn = ebnm_pn in 10 fold OCV. There is an error message

  non-finite value supplied by optim
Error in mle_normal_logscale_grad(x, s) : 
  optim failed to converge; debug information saved to temp_debug.RDS

which stops the flash code, and there is a temp_debug.RDS file the the currently working folder.

please take a look at the .rds file. (.rds is not supported, so I use zip file)

Thanks a lot.

stephens999 commented 6 years ago

this was due to very small s and x values making things numerically unstable. Fixed in 2c6fc9032d2f013c48a60e982199c4a70da868e4