Meta-analysis in PLINK - What to use in order to do a forest plot

Question

[GabrielMontenegro]

I am doing a meta-analysis using PLINK.

I have the *.assoc files for each study and I have used the --meta-analysis + qt study weighted-z options.

I have the following questions in order to do a forest plot:

1) I need the beta and SE of the meta-analysis. PLINK provides a BETA (fixed model), but not the SE. How do I compute it? I would thought I would get one, but otherwise should I use:

SE = SQRT ( 1/ SUM weights)

Where the weights = the inverse variance

2) I also need the beta and corresponding SE of each study for the forest plot. However, when I check the indvidual *.assoc files and the last columns on the *.meta files, the betas are different. Why?

[kl]

Hi,

Don't know if it's the right place but how did you convert the plink output (plink.meta) file into a format readable by R. Thought I'd ask here as you seem to have done a meta-analysis.

Thanks

[mel22]

You can import it directly in R as any txt file

[kl]

Hi,

Don't know if it's the right place but how did you convert the plink output (plink.meta) file into a format readable by R. Thought I'd ask here as you seem to have done a meta-analysis.

Thanks

Answer 1

[GabrielMontenegro]

Found an answer on the PLINK forum for my second question. The results of the beta are on the exp scale, but this was changed in a later update!

Answer 2

[shuaichengwang]

Add " --ci 0.95 " command, then it will output SE and CI.