Hotfix/issue 80 sdm likelihood evaluation explodes for extremely high kappa

[venpopov]

Summary

Set init=1 as an additional argument to the output list of configure_args from sdmSimple(). By default stan samples initial values for parameters uniformly on [-2,2], which causes sampling to fail if random effect sds on log(kappa) are initialized close to 2, because some subjects get native kappa value of > exp(15), causing the numerical integration to fail. Setting init to 1 means that Stan will sample in the range [-1,1].

I tested it with a bunch of seeds and so far it seems to have solved the problem. I am working on a custom function to easily set inits for bmmmodels for any model formula, but for now this hotfix should help with the sampling failure (which is rare though).

Added a function combine_args() to ensure that if a user provides their own inits, or any other argument they want to overwrite (except for family), the fit_model function will replace the default bmm arguments with the user ones.

Tests

[x] Confirm that all tests passed
[x] Confirm that devtools::check() produces no errors

Release notes

• Fixed a bug with sampling failure in rare cases for strange initial values for the sdmSimple() model

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (c48148e) 66.09% compared to head (8f49ae0) 66.48%.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop      #91      +/-   ##
===========================================
+ Coverage    66.09%   66.48%   +0.39%     
===========================================
  Files           14       14              
  Lines          926      937      +11     
===========================================
+ Hits           612      623      +11     
  Misses         314      314              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.