ValueError: The mean chi2 statistic of trait 1 is less than 1.02, which is too small to be well-suited for MTAG.

[cks2903]

When running MTAG I get an error message saying that the mean chi2 statistic of trait 1 is less than 1.02. What exactly is this referring to, and how can I proceed from here?
I've attached the .log-file.
OWtraitsMTAG.log

[paturley]

Hi! Sorry for the slow turn-around on this.

We've recently updated the MTAG code so it throws an error when at least one trait has a mean chi2 statistic of less than 1.02. The reason for this is that in this range, we haven't carefully tested how robust MTAG is, and there are reasons to believe the results may become a bit unstable. For example, if the mean chi2 statistic is less than one, MTAG can't even produce valid standard errors in many cases.)

If you feel comfortable with your application (e.g., if you are confident that the traits that you are using are highly genetically correlated), then we have added an option to force MTAG to produce estimates. I think we called it --force. Please correct me if I'm wrong @huilisabrina.

[cks2903]

Thank you for your answer. I am confident about the correlation, so forcing the model would be okay. However, the MTAG code doesn't seem to recognize --force.

[huilisabrina]

Hi @cks2903 ,

I was one step behind Patrick - I just added the --force option which functions as @paturley described. Could you try repulling the repo and run MTAG with --force?

Thanks,
Hui

[cks2903]

Thank you, that solved my problem!

[paturley]

Wonderful. Glad it worked.

…

On Wed, Aug 22, 2018 at 1:04 PM cks2903 ***@***.***> wrote: Thank you, that solved my problem! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#39 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AUNA9e4Tj6C0OZxpPpMR4lZAEkU57Zumks5uTY8jgaJpZM4WBba8> .

[cks2903]

I was wondering what is the mean chi2 statistic referring to?

[paturley]

The mean chi2 statistic is the mean value of the chi2 statistic (or the square of the z-statistic) across the SNPs in your data. It's equal to one (in expectation) if no SNP is associated with the trait and increases linearly in sample size otherwise. As such, it is a measure of the overall statistical power of the GWAS you are passing into the program.

…

On Thu, Aug 23, 2018 at 3:29 AM cks2903 ***@***.***> wrote: I was wondering what is the mean chi2 statistic referring to? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#39 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AUNA9dB6jA2ytTTsQ1WesGs-IZZ_cmu-ks5uTlnPgaJpZM4WBba8> .