Does DOMDEV ensure orthogonality of allele dosages and dominance deviation vectors?

[Alan Aw]

Hi,

I am trying to run some dominance deviation analyses. I noticed that PLINK provides a genotypic option under the --glm flag, which performs marginal tests of deviation from linearity across each SNP. 

In the manual, it is mentioned that the recoding for dominance deviation (DOMDEV) is 0 --> 0, 1 --> 1, 2 --> 0 (source: PLINK manual).

I was wondering if, internally, there is code that orthogonalizes the additive and dom-dev vectors prior to running the three tests. (E.g., 0 --> -p/(1-p), 1 --> 1, 2 --> -(1-p)/p would be a valid orthogonalization, as reported in Palmer et al., 2023).

I did try searching for threads related to this question, but could only find this one. Appreciate any clarification on this matter. Thank you!

Alan

[Chris Chang]

No, PLINK —glm/—linear/—logistic does not perform further recoding before regression.

--
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/cf7612ed-16fe-4b58-93c5-d1f89032ee28n%40googlegroups.com.