Standard error of the logit

[Shannon Lange]

Hi,

Does anyone know how to compute the standard error on the logit scale? I have

gen se_sens = sqrt((sensitivity - (1 - sensitivity))/n_cases)

However, this is not on the logit scale. Apparently, I cannot simply transform this variable. Any help would be appreciated!

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

I (and the others) don't really know what you're trying to do. You obviously can calculate the logit function and can use egen or summary to compute the standard error.

[Mike Lacy]

I second Phil's comments. I'm guessing, though, that each observation is a sensitivity, which is a proportion, for which the SE would be sqrt(p*(1-p)/N). Don't you want multiplication rather than subtraction in your formula? So, perhaps what you want is the formula for the standard error of a logit-transformed proportion, given the SE of the proportion?