family-GWAS: within and between-family effect, SE of the true BETA
20 views
Skip to first unread message
Gabriele Sgarlata
Oct 20, 2025, 8:30:20 AMOct 20
to plink2-users
Hi Chris,
I am performing family GWAS using the --qfam option implemented in plink1.9.
My understanding is that if I want an estimate of the direct effect (within-effect), I need to run the analysis with the option "--qfam mperm=1000 emp-se". This analysis generates two outputs:
- The *qfam.within output provides estimates of the BETA coefficient (within-effect) from the linear regression of genotype on phenotype. Since individuals are not independent (i.e., several individuals share the same family), the p-value and t-stat are to be disregarded (because they are confounded).
- The *qfam.within.perm provides i) the BETA coefficient (within-effect) from the linear regression of genotype on phenotype, ii) EMP_BETA which is the mean of the BETAs obtained from the permutation (it should be the mean of the null hypothesis), EMP_SE (which is the standard error around the sample BETA from the permutation, also representing the null-hypothesis) and EMP1 (which is the unconfounded p-value).
My understanding is thus that we cannot obtain an estimate of the standard error around the estimated BETA, accounting for the family structure, since EMP_SE represents the standard error for the BETA estimated in the permutations under the assumption of no association between genotype and phenotype. Is it correct?
Another question is: Did I understood correctly that if I want to estimate the indirect effect (between-effect) I need to perform the analysis with the "--qfam-between mperm=1000 emp-se" option?
I am performing family GWAS using the --qfam option implemented in plink1.9.
My understanding is that if I want an estimate of the direct effect (within-effect), I need to run the analysis with the option "--qfam mperm=1000 emp-se". This analysis generates two outputs:
- The *qfam.within output provides estimates of the BETA coefficient (within-effect) from the linear regression of genotype on phenotype. Since individuals are not independent (i.e., several individuals share the same family), the p-value and t-stat are to be disregarded (because they are confounded).
- The *qfam.within.perm provides i) the BETA coefficient (within-effect) from the linear regression of genotype on phenotype, ii) EMP_BETA which is the mean of the BETAs obtained from the permutation (it should be the mean of the null hypothesis), EMP_SE (which is the standard error around the sample BETA from the permutation, also representing the null-hypothesis) and EMP1 (which is the unconfounded p-value).
My understanding is thus that we cannot obtain an estimate of the standard error around the estimated BETA, accounting for the family structure, since EMP_SE represents the standard error for the BETA estimated in the permutations under the assumption of no association between genotype and phenotype. Is it correct?
Another question is: Did I understood correctly that if I want to estimate the indirect effect (between-effect) I need to perform the analysis with the "--qfam-between mperm=1000 emp-se" option?
Thank you in advance,
Best,
Gabriele
Chris Chang
Nov 15, 2025, 6:59:58 PMNov 15
to Gabriele Sgarlata, plink2-users
From the linked PLINK 1.07 documentation:
"These results are from a standard --linear type analysis, i.e. which ignores family structure. They are displayed so that the direction of effect may be determined (from the BETA) -- but otherwise, only the empirical p-value from the permuted results file should be looked at."
I.e. BETA should *not* be treated as an estimate of the direct effect, only its direction relative to EMP_BETA is informative.
"These results are from a standard --linear type analysis, i.e. which ignores family structure. They are displayed so that the direction of effect may be determined (from the BETA) -- but otherwise, only the empirical p-value from the permuted results file should be looked at."
I.e. BETA should *not* be treated as an estimate of the direct effect, only its direction relative to EMP_BETA is informative.
--
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/f1361602-99ed-4a83-90ea-26d1178135d9n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages