search

brielin / Popcorn

Software for estimating correlation of trait effect sizes across populations
33 stars 15 forks source link

both genetic-impact correlation and genetic-effect correlation greater than 1 #33

Closed gunns2 closed 1 year ago

gunns2 commented 1 year ago

Hi, After running popcorn we are finding that both the genetic-impact correlation and genetic-effect correlation are greater than 1, which was unexpected. Do you think this is the result of a bug or some error on our part? If not, how should we interpret this value?

thank you!!

brielin commented 1 year ago

hi, this is fine. the default regression model is unbounded, so it is possible to get estimates above 1. Generally the 95% confidence interval should contain 1 in this case. If you use the MLE model, the estimate will be bounded by +/- 1.

gunns2 commented 1 year ago

thank you so much!