Best model for analyzing likelihood of particular survey responses

[Tom Prosser]

Dear forum,
I'm analyzing responses to two survey questions with 4-point Likert scales and I'm interested in the demographic (i.e. age, gender, income) characteristics of respondents who responded 'definitely' to both questions. What's the best model to use? A logistic regression with a binary outcome variable? Any advice would be most gratefully received. Best and thanks,
Tom

[Maarten Buis]

A model is always a tradeoff between fit and parsimony, and what is "best" depends on how you value both goals, which typically depends on the goal of the study, intended audience, etc. What you suggest is possible, but it is up to you to decide whether it is best for your application. Here is what I think your data look like:

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(Likert1 Likert2 female)
2 3 1
. 4 0
2 4 1
4 3 1
1 3 1
4 3 0
4 3 0
4 4 1
4 4 0
4 2 0
4 4 1
4 4 1
4 3 0
2 4 1
3 4 0
4 2 0
2 3 1
4 3 0
4 4 0
1 3 0
4 4 0
3 4 1
3 4 0
4 3 0
4 3 0
4 4 0
2 2 0
3 3 0
2 4 1
2 4 0
end
label values Likert1 Likert_lb
label values Likert2 Likert_lb
label def Likert_lb 1 "definately not", modify
label def Likert_lb 2 "probably not", modify
label def Likert_lb 3 "probably", modify
label def Likert_lb 4 "definately", modify
label values female sex_lb
label def sex_lb 0 "male", modify
label def sex_lb 1 "female", modify

This is one way to create a variable that is 1 when both Likert variables are 4, 0 when one or both are less than 4, and missing when one or both of them are missing:

gen both = 0 if !missing(Likert1, Likert2)
replace both = 1 if Likert1 == 4 & Likert2 == 4

After that you can use that variable in a logit:

logit both i.female, or

[Tom Prosser]

Thanks Maarten - that's really useful and very kind of you, Tom