Lowest limit of rg p-value reported?

[jodithea]

Hi,

The output from running a genetic correlation (--rg) has given 'P: 0.' I am guessing this may be because there is a lower limit to what P-value can be reported. I was wondering what this lower limit is for reporting purposes - I could report P < lower limit (e.g. P < 1e-300). Another genetic correlation I have run has given 'P: 5.9509e-235' so it must be something below that.

Thanks.

[aksarkar]

@jodithea The limit is the numerical precision of scipy.stats.chi2(1).sf.

Those p-values are suspiciously small, suggesting something has gone wrong. What are the z-scores?

[jodithea]

Thanks for your reply @aksarkar. The output is:

Genetic Correlation

Genetic Correlation: 0.9507 (0.0214)
Z-score: 44.3288
P: 0.

and

Genetic Correlation

Genetic Correlation: 0.9144 (0.0279)
Z-score: 32.7295
P: 5.9509e-235

I am expecting a very high correlation as I am correlating the results from my own GWAS to the largest publically available GWAS of the same trait.

[aksarkar]

I suggest reporting the confidence interval or the z-score for the genetic correlation without the p-value.

[jodithea]

OK fantastic, thanks so much for your help.
One last question - is the value in brackets the standard deviation or the standard error? (i.e. 0.0279 from 'Genetic Correlation: 0.9144 (0.0279)'). Sorry if this is obvious, I've struggled to find info on this in the documentation.

[aksarkar]

@jodithea It's the jacknife standard error.

[jodithea]

Thanks so much for your help @aksarkar