Where can I find the factor loading estimates of userGWAS and commonfactorGWAS?

[roya karimi]

Hello Andrew and everyone else who is reading this post,

I have a basic question about the graph in the tutorial. I understand that the plots in the GitHub tutorial are created manually using the outputs of the analysis. My question is, where did you get the values for the estimates of F1=~ MDD+PTSD+ANX+ALCH? While I can find the F1~SNP estimates and their ES in the output table of userGWAS and commonfactorGWAS, I am unsure of where to find the values for the F1 and its indicators estimate. I have attached a photo of the graph and marked the values that I am unable to find here.

Any help would be greatly appreciated.

Best,

Roya

[agro...@gmail.com]

Hi Roya, 

Great question! For pragmatic reasons, commonfactorGWAS only saves the SNP effects on the factor. Similarly, we always recommend using the 'sub' argument to specify which specific effects you want to save from the model (e.g., F1~SNP) so that you are not saving the full model output for every single SNP. The primary reason being that it will seriously increase the memory needs for the job if you save every single model parameter for each SNP. In addition, we not expect the parameters you highlighted in yellow to change very much for each SNP so it's often safe to assume that these values are approximately equivalent to the estimates you get when running the model using usermodel for the model without the individual SNP effects. 

With all that said, I just put in a recent update to userGWAS this week to include a fix_measurement argument that tells the function to fix all of these estimates to improve interpretation, as each SNP can now be interpreted as going through the exact same measurement model. Their may be some instances where you want to see if the measurement model shifts for a particularly large effect SNP, in which case you can set fix_measurement to FALSE and not use the sub argument to get all of the model parameters. I'll be documenting this new fix_measurement argument in the coming weeks on the wiki, but hopefully this more generally answers your question for now. 

Best,

  Andrew

[roya karimi]

Thank you very much Andrew for your response. Your explanation has clarified my doubts. Looking forward to the new update. Thank you again.