Testing Discrete distributions

[mstasinopoulos]

The idea is twofold

i) to make sure that the distribution provide the right results for values outside the range
ii) gamlss and gamlss2 produce compatible results

The function to test this are

i) test_continuous_gamlss_dist() and
ii) test_discrete_gamlss_dist()

I am attaching the two folders here

test_discrete_gamlss_dist.txt

test_continuous_gamlss_dist.txt

The results for testing the discrete distribution cases can be summarises as follows

The distributions GEOM , GEOMo, LG, PO, YULE, ZIPF, NBI, NBII, PIG, WARING, ZALG ZAP, DEL, SI, SICHEL ,
ZANBI, BI,BB, ZABI, ZIBI, ZIBB are OK.

DPO, PIG2, BNB, DBURR12 NBF, ZAPIG, ZINBI, ZIPIG, ZABNB , ZASICHEL , ZIBNB , ZISICHEL, DBI need checking

ZIP and ZIP2 produce different deviance in gamlss and gamlss2

PSPIG need impementing

Mikis

[zeileis]

Thanks! Reto @retostauffer (and some discussion with me) has started implementing tinytest suites that can test (a) the distributions and (b) the family objects. This will re-use many of the strategies from your scripts above.

The consistency between gamlss and gamlss2 should be tested elsewhere, e.g., in gamlss2 as in gamlss-dev/gamlss2#12

[retostauffer]

@zeileis what we could also discuss is a dedicated repository containing test-scripts to compare gamlss and gamlss2 not to clutter the repositories for the packages themselves. TBD.