Suppose I have two Independent Variables (A, B), and a three-level categorical dependent variable (WOM valence, 1=negative, 2=mixed, 3=positive).
I am interested in the difference in the percentage of negative/mixed/positive WOM between B1, B2, under each level of A. In other words, I want to compare the difference between a1b1% and a1b2% for negative, mixed, and positive, separately. And also the difference between a2b1% and a2b2% for negative, mixed, and positive, separately.
IVs: A (a1, a2) x B (b1, b2)
DV: three-level categorical variable (1=negative, 2=mixed, 3=positive)
Is there any approach to analyze this data by SPSS or Stata using multinomial logistic regression?
Thanks in advance!
With paired data you either want a multilevel logistic regression or something similar like GEE (generalized estimating equations). SPSS and State can handle either approach (though I'm an R user these days). Also I wonder if you'd be better off with an ordinal logistic regression?
Thom Baguley Thanks Thom! It's very insightful. Could you please tell me where can I find some tutorials about multilevel logistic regression or GEE? I never used anything like those before. It's a little beyond my reach though.
Not off hand - but UCLA have lots of online examples that might be a good place to start.
Pete Pengcheng Zhou, In Stata you will need to use the -gsem- command to estimate a multilevel multinomial logistic regression model. See Example 41g in Stata's documentation for an application of this method.
You can follow up and read in the research
ستخدام الانحدار اللوجستي متعدد الاستجابة لتحديد العوامل المؤثرة على مرض العيون
- February 2019
- Projects:
- الانحدار اللوجستي المتعدد
- RG Achievement🌷
- 📷Haider Raid Talib
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs