Plink genotype * sex

Question

[rmretama]

Hi,

I'm doing logistic regression analysis and I would like to test if there is a genotype and sex interaction.

I did this analysis:

plink1.9 --bfile BFILE --extract SELECTED_variants --logistic genotypic sex interaction --adjust --pheno BFILE_all-pheno --all-pheno --ci 0.95 --out Regression_logistic+geno+sex+interac

So, I wondered if this was correct, since I read that "interaction" evaluates a genotype x covariates interaction. Then I did this analysis:

plink1.9 --bfile BFILE --extract SELECTED_variants --logistic genotypic sex interaction --covar PCA.evec --covar-name PC1 PC2 --adjust --pheno BFILE_all-pheno --all-pheno --out Regression_logistic+geno+sex+covar+interac

I'm new on this logistic regression analysis, I would really appreciate if somebody could check if I'm doing this right.

Thanks!

[Ram]

Please use the formatting bar (especially the code option) to present your post better. I've done it for you this time.

Answer 1

[chrchang523]

(Cross-posting this from plink2-users.)

--logistic's 'interaction' modifier causes all genotype-covariate interaction terms to be included in the regression. If you only want to include the genotype x sex interaction term, and not the genotype x PC1, genotype x PC2, ... terms, you'll need to use the --parameters flag to select the regression terms you're keeping; see the --parameters section of https://www.cog-genomics.org/plink/1.9/assoc#linear for more details. Yes, working out the correct numbers for --parameters is annoying.

'genotypic' should only be used if you want an extra "dominance deviation" genotype column; it doesn't sound like this is the case.