big change for rg and gencov after using --no- intercept

[Fei Yangf]

Hi ldsc developer, 

After using no-intercept in our case to calculate the genetic correlation, we found the big changes for gencov (from negative to positive) and rg (from negative to positive, its p value also became non-significant to significant). Because we are pretty sure there is no case sample overlap between two datasets, maybe only a few control sample overlap.  And we also used Tetrachoric correlation to detect the sample overlap between two datasets, the result is 0.015. We don't know why these two datasets got  so different results whether constraining the intercept or not.  

By the way, we also calculate the genetic correlation of each trait above with other trait, no big changes for rg and gencov were found using no-intercept or not. It seems that we can get big change for rg and gencov only when calculate the genetic correlation between those two traits above. I attached four result log files. Really hope that it can be solved!  

Thanks, 

Fei