topic Re: Two parameter estimates and OR in Discussions

Two parameter estimates and OR

jmplearner2 — Fri, 17 May 2024 23:01:47 GMT

I am not sure why I have two parameter estimates outputs for the variables after it finish running it once - starts again with a new estimate and gives different estimates. I am running a nominal logistic regression with a dependent variable that has 3 choices (multi-variable).

Also, I can't seem to have the choice to have OR calculated and only gives me the option for the confidence interval and Wald's test.

Would I calculate the OR by doing Exp of the estimate? or regression co-efficient (I am a bit rusty on my stats).

Please help. Thanks in advance

Re: Two parameter estimates and OR

dlehman1 — Fri, 17 May 2024 23:26:55 GMT

I'm not sure I understand your first question, but if you response variable has multiple levels, you should have multiple parameter estimates table that relate your factors to the different levels. I believe JMP only provides the odds ratios for a binary response variable. But if you save the prediction formula from the logistic regression, it will give you the probabilities. You can then tabulate these for the groups you want to compare and calculate the odds ratios from the probabilities.

Re: Two parameter estimates and OR

jmplearner2 — Sat, 18 May 2024 19:23:00 GMT

Thank you. That is helpful.

Also, can you help me interpretate the figure below. The level1/level2 is throwing me off.

I changed the DV to binary (heavy cannabis users = 1 vs light cannabis user = 0) and those are age ranges (e.g. 26-35year old, and 21-25 year old).

So for the first line, would the right interpretation be that compared to 21-25 years old's, 26-35yo have 1.629 times the odds of being a heavy cannabis user?

Thanks in advance.