paturley
commented
2 years ago Hello,
In general, the main things that affect the mean chi2 statistic of a GWAS is the heritability of the phenotype and the sample size available. So other than considering other more heritable phenotypes or gathering more data, I'm not sure that there are valid ways of boosting the chi2. Generally, I don't recommend running MTAG for phenotypes where the mean chi2 is less than 1.02 since MTAG's SEs become quite unreliable and the risk of an inflated Type I error increases too much in that range.
MTAG should work well with both binary and continuous phenotypes as long as the mean chi2 is large enough and MTAGs assumptions are met.
Sorry I can't be of more help.
Best, Patrick
On Mon, Sep 19, 2022 at 3:21 PM nilaygun @.***> wrote:
Hello,
For some traits included in joint analysis, I get a mean chi2 statistic smaller than 1.02. If my understanding is correct, this value is equal to the mean of squared z-score values in the sumstat file; therefore it is only contingent upon beta estimate and error. May I ask your advice on which parameters I can tune in my GWAS data prior to running MTAG to obtain z-scores that would give appropriate mean chi2 statistics to run MTAG (e.g. normalization of phenotype, number of features in covariate file)? Or in general, would you recommend not applying MTAG for the traits when the mean chi2 statistic is smaller than 1.02? Also, in my GWAS analysis, I evaluate a continuous variable as a phenotype trait rather than a binary outcome (i.e. disease), may I also ask if MTAG would be applicable for GWAS with continuous traits?
Many thanks in advance for your help!
— Reply to this email directly, view it on GitHub https://github.com/JonJala/mtag/issues/170, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFBUB5JPQ7F42RWC4NTQLILV7C4MTANCNFSM6AAAAAAQQMQS3A . You are receiving this because you are subscribed to this thread.Message ID: @.***>
nilaygun
Hello,
For some traits included in joint analysis via MTAG, I get a mean chi2 statistic smaller than 1.02. If my understanding is correct, this value is equal to the mean of squared z-score values in the sumstat file; therefore it is only contingent upon beta estimate and error. May I ask your advice on which parameters I can tune in my GWAS data prior to running MTAG to obtain z-scores that would give appropriate mean chi2 statistics to run MTAG (e.g. normalization of phenotype, number of features in covariate file)? Or in general, would you recommend not applying MTAG for the traits when the mean chi2 statistic is smaller than 1.02? Also, in my GWAS analysis, I evaluate a continuous variable as a phenotype trait rather than a binary outcome (i.e. disease), may I also ask if MTAG would be applicable for GWAS with continuous traits?
Many thanks in advance for your help!