Post by emilowk on Apr 12, 2016 6:56:38 GMT
Is it possible to correct the GCTA-GREML estimates for measurement error?
Gwern used the standard attenuation formula (https://en.wikipedia.org/wiki/Correction_for_attenuation), but I'm not sure this is the correct to use with GREML results.
In particular, the Davies et al 2016 paper found a GREML-based h2 of .31. However, the test they used had a test-retest coefficient of only .65, so there is a substantial amount of measurement error (r manifest variable x latent variable only = .81 (sqrt(.65)). How does one correct the value?
Gwern used the standard attenuation formula (https://en.wikipedia.org/wiki/Correction_for_attenuation), but I'm not sure this is the correct to use with GREML results.
In particular, the Davies et al 2016 paper found a GREML-based h2 of .31. However, the test they used had a test-retest coefficient of only .65, so there is a substantial amount of measurement error (r manifest variable x latent variable only = .81 (sqrt(.65)). How does one correct the value?