How is the p-value in --glm (GWAS) calculated?

[Monica Isgut]

Hi,

Just a quick question - it is explained that the --glm function, for binary categorical data, is a univariate logistic regression algorithm that sequentially trains on each SNP. How does Plink calculate the p-values afterward (ex. chi-squared test)?

Thanks!

[Christopher Chang]

Plink's logistic regression p-value is based on a Wald test.

[Ziang Zhang]

Hi,

I would like to quickly follow up on this question - if we use the --glm to fit a logistic regression but with the 'genotypic' modifier, is the p-value of the 2 df joint test (e.g. "GENO_2DF") still computed based on the Wald test? 

Thank you!

[Christopher Chang]

From the 'genotypic' modifier documentation (https://www.cog-genomics.org/plink/2.0/assoc#genotypic ):

The 'genotypic' modifier adds a dominance-deviation column (heterozygous-A1 = 1, any other genotype = 0, linear interpolation applied to dosages; "0..1..0" for short) to G, and adds genotype + dominance-deviation joint F-test results to the main report; the test name is "GENO_2DF".

[Ziang Zhang]

Thank you for your response. I understand that "GENO_2DF" corresponds to a joint F-test when a linear regression (instead of logistic regression) is fitted. For --glm genotypic on a binary trait (which fits a logistic regression), how is that 2 df GENO_2DF test done? Is it using Wald, LRT or score test?

[Chris Chang]

Wald test, where the p-value is computed from the appropriate F distribution.

--
You received this message because you are subscribed to the Google Groups "plink2-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plink2-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/plink2-users/d95bdb91-8128-4e05-ab8e-f96987567acan%40googlegroups.com.