Subscribe Bookmark RSS Feed

calculate odds ratios for several predictors

Thaelina

Occasional Contributor

Joined:

Dec 13, 2016

My data consists of two predictors, one with 3 responses (time at exposure) and 1 with two responses(gender.

I can do a log reg just fine and I cross my two variables, but I'm not sure how to interpret my OR's, .

I want to know: OR for early exposed females. How do I get that? Is that already in the table, or do I need multiply the OR for females with the OR for early exposed?

And if it's already in the table, how do I get early exposed males?

 

I've talked with my advisor and he insists that nominal is the way to go, so that's what I have done

1 ACCEPTED SOLUTION

Accepted Solutions
dale_lehman

Community Trekker

Joined:

Jan 29, 2015

Solution

You are correct - it appears that JMP does not provide odds ratios for the interactions.  Further, it seems to be somewhat complex to calculate them, but it can be done.  I found the following to provide some detail:

http://www.ats.ucla.edu/stat/stata/seminars/interaction_sem/interaction_sem.htm

Also, the early slides in

http://www.cantab.net/users/filimon/cursoFCDEF/will/logistic_interact.pdf

appear to show what to do.

4 REPLIES
dale_lehman

Community Trekker

Joined:

Jan 29, 2015

I might be wrong, but is sounds like you should include as a predictor the cross between time and gender.  It will have its own odds ratios which is what you are looking for.  It is entirely possible that you might be able to derive that from the odds ratios for the two predictors alone (although I doubt it), but I'm not smart enough to know how to do that.

Thaelina

Occasional Contributor

Joined:

Dec 13, 2016

But I think I already did that? I have 3 model effects: Gender, Exposure status and gender*exposure status, and I don't get any ORs for that. I get parameter estimates and effect likelihood ratio for it though, that's why I thought I might be able to calculate it. It might also be a part of the given OR's I just can't figure it out.

 

This is the OR output, but where is female-late neutering against unaltered?

 

Level1/Level2Odds RatioProb>ChisqLower 95%Upper 95%
Late neuteringEarly neutering2.85591490.19240.589562313.834415
UnalteredEarly neutering1.22802270.74890.34906454.3202328
UnalteredLate neutering0.42999270.12030.14826361.2470607
Early neuteringLate neutering0.35015050.19240.07228351.6961735
Early neuteringUnaltered0.81431720.74890.2314692.8648004
Late neuteringUnaltered2.32562070.12030.80188566.744742
      
Odds Ratios for Gender     
Level1/Level2Odds RatioProb>ChisqLower 95%Upper 95%
MaleFemale1.15815070.7890.39506573.3951648
FemaleMale0.86344550.7890.29453652.5312246

 

 

 

dale_lehman

Community Trekker

Joined:

Jan 29, 2015

Solution

You are correct - it appears that JMP does not provide odds ratios for the interactions.  Further, it seems to be somewhat complex to calculate them, but it can be done.  I found the following to provide some detail:

http://www.ats.ucla.edu/stat/stata/seminars/interaction_sem/interaction_sem.htm

Also, the early slides in

http://www.cantab.net/users/filimon/cursoFCDEF/will/logistic_interact.pdf

appear to show what to do.

Thaelina

Occasional Contributor

Joined:

Dec 13, 2016

Oh dear,
Well, thank you so much for those links, they're exactly what I need!