topic Re: Obtaining 95% CI from predicted P (logistic) in Discussions

Obtaining 95% CI from predicted P (logistic)

fishguy — Fri, 09 Jun 2023 00:53:47 GMT

Hello,

I'm doing logistic regression in JMP 16. I would like to obtain the 95% CI around predicted P for some independent variables using the P as a measure of strength of association. I know the math to calculate the 95% CI from linear predictors but I need help pulling out the necessary information once I run the model. I can obtain the linear predictor by saving to a new column, but I also need the standard error of the linear predictor. How do I obtain that? I have done this in R but I would prefer to learn in JMP.

Thanks!

Re: Obtaining 95% CI from predicted P (logistic)

pauldeen — Wed, 17 Aug 2022 07:46:03 GMT

do the logistics modeling use Fit model you can turn on CI in the hotspot menu:

This will give you CI on the parameter estimates table.

From the same hotspot you can also save the probability formula's.

Re: Obtaining 95% CI from predicted P (logistic)

fishguy — Wed, 17 Aug 2022 17:11:48 GMT

Thanks for reply. Turning on CI gives me the CI around the chiSq tests for each term, but what I'm after is obtaining a CI around the expected value (probability of "1" or "yes"). To calculate the CI, one needs the linear predictor as well as the SE. I can get the linear predictor and the expected value from the save drop down choices but I can't find where to obtain the SE of the linear predictor.

95% confidence interval formula:

Lp-1.96* se(Lp), Lp+1.96*se(Lp)

Lp - linear predictor

Re: Obtaining 95% CI from predicted P (logistic)

fishguy — Wed, 17 Aug 2022 17:16:09 GMT

...and to finish the thought, this is the formula to get CI around probabilities (expected values) once you have the linear predictor interval:

Lp interval [a, b]

Probability interval:

[exp(a)/1+exp(a), exp(b)/1+exp(b)]

[lower, upper]

Re: Obtaining 95% CI from predicted P (logistic)

pauldeen — Thu, 18 Aug 2022 08:43:01 GMT

Did you notice the Std Error column in the parameter estimates table?

Re: Obtaining 95% CI from predicted P (logistic)

fishguy — Thu, 18 Aug 2022 18:44:46 GMT

Thanks much! My bad. I thought it was for something else.

Re: Obtaining 95% CI from predicted P (logistic)

pauldeen — Fri, 19 Aug 2022 06:24:42 GMT

Cool! If you are satisfied, please mark the topic as solved so that other people know it is done :)