Analysing multiple response variables

[Vincent Kayemba]

I want to analyse the relationship (using Chi-square test) between (AgeNEW, Gender) Vs (Q42-production problems people encounter in pigeon pea production). The challenge is that Q42 is a Multiple response question whose options are broken down in variables Q42a-Q42m. Is there a away i can generate a single variable capturing the variables Q42a-Q42m as categories under the new variable?
If yes, how do i go about it?
If no, please guide me on how best i can analyse this multiple response question.

[Carlo Lazzaro]

Vincent:
I would take a look at -help pca-.

[Nick Cox]

#1 sounds contradictory to me. The whole point of a multiple response question is that different answers are possible -- and even that two or more different answers are possible. Take for example a question Which of the following do you use in your statistical research?

Stata
SPSS
SAS
R
Python
MATLAB
Julia
Excel
...
Other

(apart from Other, the ordering is naturally just reverse alphabetical).

What kind of reduction to one dimension would even make sense here? Perhaps the number of positive responses. Perhaps the most important axis of the binary matrix as implied by PCA (Carlo #2) or correspondence analysis, but good luck with getting an easy interpretation. With the latter your goal of chi-square testing still implies the need for a reduction to categories, not a measure.

(Your question is clearly different but I lack the specialist knowledge to discuss it.).