paturley
commented
1 year ago Hi Allen,
MTAG also uses the N as a proxy for the SE^2 in the calculation. If you are pretty confident that the sample size is constant across SNPs, you can get a pretty good estimate of the appropriate value for N if you use
N = 1/(2p(1-p)*SE^2)
where p is the allele frequency of the SNP and SE is the standard error for the SNP. A heads up though that the betas and SEs that MTAG reports may be in different units than the units of the GWAS. One way to verify how the units differ is to pass each of the sets of sumstats through MTAG one at a time. The new betas should be a constant multiple of the old betas, which should inform how you can transform the MTAG betas into the original betas if you like. Even if you don't transform the betas and SE though, the Z-stat and p-values should be correct.
On Thu, Sep 7, 2023 at 10:48 AM allenXY @.***> wrote:
Hi,
A great project! Currently I have some questions about the column "n" in the input of sumStats files: (1). in paper, you mentioned "we restrict variation in SNP sample sizes by calculating the 90th percentile of the SNP sample size distribution and removing SNPs with a sample size smaller than 75% of this value". Is it the only place where MATG uses the information of "n" for each SNP? Are there any other places MTAG uses these "n"?
(2). if the input sumStats files do not have such "n" column, how should we generate it by ourselves? I noticed that there are different answers for this question in ISSUES, so I am confused. Specifically, if I have three different studies: (a). case-control study with binary outcome; (b). quantitative study with continuous outcome; (3). survival study with (time, disease status) outcome, then how should we generate the "n" column for these three studies.
Thanks, Allen
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/185, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5OY2O7VM2HWFD4RANDXZHNEFANCNFSM6AAAAAA4PAZYN4 . You are receiving this because you are subscribed to this thread.Message ID: @.***>
allenXyGao
dqq0404
Hi,
A great project! Currently I have some questions about the column "n" in the input of sumStats files: (1). in paper, you mentioned "we restrict variation in SNP sample sizes by calculating the 90th percentile of the SNP sample size distribution and removing SNPs with a sample size smaller than 75% of this value". Is it the only place where MATG uses the information of "n" for each SNP? Are there any other places MTAG uses these "n"?
(2). if the input sumStats files do not have such "n" column, how should we generate it by ourselves? I noticed that there are different answers for this question in ISSUES, so I am confused. Specifically, if I have three different studies: (a). case-control study with binary outcome; (b). quantitative study with continuous outcome; (3). survival study with (time, disease status) outcome, then how should we generate the "n" column for these three studies.
Thanks, Allen