Question

Function Rf_lgammafn() in the log_posterior function in the DESeq2 C++ functions

0

Entering edit mode

vtshen • 0

@vtshen-13292

Last seen 7.7 years ago

Hi,

I am learning the DESeq2 package. I have derived the log likelihood function in the negative binomial distribution, that is,

In the log_posterior function, the log likelihood part is

ll_part =  sum(lgamma(y + alpha_neg1) - Rf_lgammafn(alpha_neg1) - y * log(mu + alpha_neg1) - alpha_neg1 * log(1.0 + mu * alpha));

I am not understanding this term above: Rf_lgammafn(alpha_neg1). I found that the function lgammafn() computes log|gamma(x)|, but I am thinking to use lgamma(alpha_neg1).

Thanks

Shen

deseq2 • 1.1k views

ADD COMMENT • link updated 7.8 years ago by Michael Love 43k • written 7.8 years ago by vtshen • 0

score 1 · Accepted Answer · 2017-06-20

1

Entering edit mode

Michael Love 43k

@mikelove

Last seen 3 days ago

United States

I'm sorry, I don't follow the question. The code and formula seem to match (reordered). Are you asking why Rf_lgammafn instead of lgamma? Of course you can reimplement the log likelihood however you like :)

ADD COMMENT • link 7.8 years ago Michael Love 43k

0

Entering edit mode

Hi Prof Love, thanks for the quick reply. My question was not quite clear but I think you got it right. The question is why Rf_lgammafn instead of lgamma? Based on the logL formula, my first thought is to use lgamma() for the term log gamma(alpha_neg1), but you used Rf_lgammafn(). What is the difference between lgamma() and Rf_lgammafn()? Thanks.

ADD REPLY • link 7.8 years ago vtshen • 0

0

Entering edit mode

I wrote this a while ago (2012) but I believe it's because alpha_neg1 is a double and the other quantities are Rcpp objects

ADD REPLY • link 7.8 years ago Michael Love 43k