Using powerlog command to calculate power for case-control study

[Serena Yue]

I am looking at a blood biomarker (continuous)'s ability to predict gestational diabetes (binary outcome). I have 208 women without the outcome (controls), and 56 women with the outcome (cases). I wish to know what is my power for detecting a meaningful difference of an Odds Ratio of 1.5. Importantly, the blood biomarker is protective of the outcome, i.e. women with the outcome have lower levels of the biomarker.

According to the STATA help manual for powerlog, it says that "p2 = estimated probability of a 1 at the mean + 1 sd of the predictor", does this assume that there is a positive association between the predictor or outcome? I am wondering whether I should calculate p2 as the mean minus 1sd of the predictor in my case?

I also am not sure how/where to factor in the OR 1.5 I am trying to detect..

Lastly, how can I input the sizes of my control and cases group, e.g. can I type:

powerlog, p1(x) p2(y) n1(56) n2(208)?
or powerlog, p1(x) p2(y) n1(208) n2(56)?

I am using STATA version 16. The powerlog command is externally installed.

Thank you so much for everyone/anyone's help!

[Rich Goldstein]

there is much that is unclear about what you are doing but I think the following might help at least a little: (1) the logistic function is symmetrical so mean+1sd is the same, in the power situation, as mean-1sd; (2) there are not options to this command allowing you to input your n's; as per the FAQ, you should have mentioned that -powerlog- is user-written; one of the things that is not clear is exactly what you are trying to do: at first it appears you want to calculate the needed sample size, but then it appears you want the effect size discoverable with a particular sample size; please read the FAQ, which has advice on asking good questions, and re-post