Because your dependent variable has two levels, this is a relatively tractable situation. You can use (binary) logistic regression. Variable 1 will be treated as dummy variables. For Variable 2, I think you could handle it two ways: a) Treat each possible choice as its own binary variable (yes/no). Note that this is different than representing it as dummy variables. b) Treat it as a count. (How many options did the respondent choose.)

Thank you so much for your answer!

Good morning Danielle Reed. I agree with the advice Sal Mangiafico gave you. But I don't understand what the x 6 and x 2 meant in your descriptions of variables 3 and 2:

categorical variable 1: (level 1, level 2, level 3) x 6

categorical variable 2: (level 1, level 2, level 3, level 4, level 5, level 6) x 2

Can you clarify?

Meanwhile, here are some resources on how to estimate logistic regression models via SPSS that you may find helpful.

• https://www.theanalysisfactor.com/spss-procedures-logistic-regression/
• https://www.restore.ac.uk/srme/www/fac/soc/wie/research-new/srme/modules/mod4/index.html
• https://stats.oarc.ucla.edu/spss/dae/logit-regression/
• https://stats.oarc.ucla.edu/spss/output/logistic-regression/

Bruce Weaver , I think you're just reading it wrong.

The x goes with the variable following it. So x 6 : Variable 2 has 6 levels.

So the overall contingency table is 3 x 6 x 2 .

Good morning Bruce Weaver . Now that you have mentioned it, I just realised my mistake with my analysis/design statement. Maybe I should have said "A 3 (variable 1: level 1, level 2, level 3) by 6 (variable 2: level 1, level 2, level 3, level 4, level 5, level 6) between-subjects design. The dependent variable (or outcome) is variable 3 with levels 1 and 2 ". Thanks for picking that up, I am still new to research (doing honours).

Thank you for the resources!

Ah, yes! Thanks Sal Mangiafico. Maybe I need another coffee? ;-)

Thank you Sal Mangiafico for explaining what I meant

I guess it depends on the style for where you are publishing, but I find either description of the design pretty difficult to follow. How about:

A 3 x 6 x 2 between-subjects design, where Variable 3 was treated as the dependent variable (with two levels, Level 1 and Level 2). Variable 2 (6 levels, multiple response) and Variable 1 (3 levels, multiple choice) were treated as independent variables.

If need be, you could add additional information about Variable 2 and Variable 1 in another sentence.

People in psychology (or at least the ones I know) would understand 3 x 6 x 2 to be describing 3 explanatory variables. So for psychology journals & audiences, you would have to change that to a 3 x 6 binary logistic regression model. But 3 x 6 would likely [EDIT: added would likely] be taken to mean that the interaction of the two explanatory variables was included in the model. If you did not include the interaction, it would be clearer to say that you estimated a binary logistic regression model with Y as the DV (perhaps with a brief description of the two values), and with A (3 levels) and B (6 levels) as the explanatory variables. For absolute clarity, you could state that the interaction was not included in the model. But presumably, your report will also include a table showing regression coefficients, SEs and ORs. If so, readers will be able to see which terms were (and were not) included in the model. This is my opinion, of course. YMMV. HTH.

Yeah, I did not mention it but I am also looking at the interaction between variable 1 and 2 on variable 3. That would have been good contextual information to add to my question... oh well.

Thank you for your input :)

Thank you Sal Mangiafico for the design statement suggestion