Complaints over gcf

[jinghuazhao]

I appeared to come across a fixable issue to do with the Numerical Recipes routine. My log says

FATAL ERROR - a too large, ITMAX too small in gamma countinued fraction (gcf)

so the -1.info file could not be generated.

I replaced several routines in MathStats.cpp with those from the web, http://numerical.recipes/forum/attachment.php?attachmentid=60&d=1190409664 (I suspect its gamp should have been gammp), and the error went away.

Otherwise, I have also seen a cleaner one here, http://www.if.ufrj.br/~bertu/incomplete_gamma.cpp, but didn't experiment with it.

[shugamoe]

Just to confirm, this error will not cut off the writing of the result *.txt file, right?

I think with VERBOSE ON it prevents the writing of both the *.txt and *.info files, but without VERBOSE it seems to write the same number of lines every time to the *.txt file so I think it shouldn't cut off the *.txt writing, but that number seems smaller then what I expect.

[shugamoe]

It seems I was reaching this error by not using METAL to log the OR column, so no worries on my part. Thank you for your continued work on this!

[jinghuazhao]

I just got to see notification from GitHub. It has been a while but I could replicate the error and the .tbl file is only partially generated (230M ~ 1.3GB in our data). The METAL script I used is copied here.

SEPARATOR TAB COLUMNCOUNTING STRICT CHROMOSOMELABEL CHR POSITIONLABEL POS CUSTOMVARIABLE N LABEL N as N TRACKPOSITIONS ON AVERAGEFREQ ON MINMAXFREQ ON ADDFILTER AF1 >= 0.01 ADDFILTER AF1 <= 0.99 MARKERLABEL SNP ALLELELABELS A1 A2 EFFECTLABEL BETA PVALUELABEL P WEIGHTLABEL N FREQLABEL AF1 STDERRLABEL SE SCHEME STDERR EFFECT_PRINT_PRECISION 8 STDERR_PRINT_PRECISION 8 GENOMICCONTROL OFF LOGPVALUE ON OUTFILE BROX- .tbl PROCESS 1-BROX.fastGWA.gz PROCESS 2-BROX.fastGWA.gz PROCESS 3-BROX.fastGWA.gz ANALYZE HETEROGENEITY CLEAR

[xingejun]

Hi,

I appeared to come across a fixable issue to do with the Numerical Recipes routine. My log says

FATAL ERROR - a too large, ITMAX too small in gamma countinued fraction (gcf)

so the -1.info file could not be generated.

I replaced several routines in MathStats.cpp with those from the web, http://numerical.recipes/forum/attachment.php?attachmentid=60&d=1190409664 (I suspect its gamp should have been gammp), and the error went away.

Otherwise, I have also seen a cleaner one here, http://www.if.ufrj.br/~bertu/incomplete_gamma.cpp, but didn't experiment with it.

Hi, jinghua

I met this error these days, too. Could you mind helping me to solve it. For example, could you mind share your MathStats.cpp code file with me? Sorry, I could not very understand how to replace routine code because I seldom used C.

Thank you very much.

Xinger

[jinghuazhao]

Hi Xinger,

I am short of making a pull-request for METAL since I have made a revised direction field in our meta-analysis which may cause confusion to others. A trade-off is to make the MathStats.cpp separately as suggested,

The copy can be found here, https://github.com/jinghuazhao/tests; I suppose you would need click METAL/, then MathStats.cpp, and Raw to have the source.

Please let me know if you see anything odd,

Jing Hua