rkwalters
commented
6 years ago Hi Carolina, This looks similar to a previously reported bug with liability h2 values in the rg results table. Thought it was fixed, but this looks like it's still an issue. (Tim, I think this is something you had looked into at some point?)
Assuming it's the same bug, the values earlier in the log are correct. The values in the table are incorrect due to using the wrong prevalence for liability scale conversion. If you want to verify, try running ldsc without the prevalence arguments to get all observed-scale estimates and apply the liability scale conversions manually.
Cheers, Raymond
On Jan 8, 2018 4:37 PM, "croselli" notifications@github.com wrote:
Dear ldsc team,
I am running -rg on 11 traits with conversion to the liability scale by including population and sample prevalences. When inspecting the output file, I observed a discrepancy between the "Total Liability scale h2" listed in the output file on top, versus the "h2_liab" and "h2_liab_se" in the "Summary of Genetic Correlation Results" table. The values are always identical for phenotype 2 (first row of table), even if I change the order of the traits. But for all subsequent phenotypes there are strong deviations.
For example: Total Liability scale h2: 0.1237 (0.0171) versus: h2_liab h2_liab_se 0.5629 0.0780
Based on the description of the results file, I was expecting the values to be the same in the results table, but I might be misunderstanding the output. Could you explain the reason for these discrepancies? And between those values, what is the correct heritability estimate on the liability scale?
Thank you, Carolina
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/bulik/ldsc/issues/99, or mute the thread https://github.com/notifications/unsubscribe-auth/AILEvUWvLvd4JyjnqYdTr4xrgJWX7yDCks5tIoqbgaJpZM4RW_So .
croselli
Dear ldsc team,
I am running -rg on 11 traits with conversion to the liability scale by including population and sample prevalences. When inspecting the output file, I observed a discrepancy between the "Total Liability scale h2" listed in the output file on top, versus the "h2_liab" and "h2_liab_se" in the "Summary of Genetic Correlation Results" table. The values are always identical for phenotype 2 (first row of table), even if I change the order of the traits. But for all subsequent phenotypes there are strong deviations.
For example: Total Liability scale h2: 0.1237 (0.0171) versus: h2_liab h2_liab_se 0.5629 0.0780
Based on the description of the results file, I was expecting the values to be the same in the results table, but I might be misunderstanding the output. Could you explain the reason for these discrepancies? And between those values, what is the correct heritability estimate on the liability scale?
Thank you, Carolina