How to Model Multiple Responses?

Dear All,

I need your assistance in choosing the right analysis method. A sociological research came to see if there are relations between people’s ethnicity and their opinion on who is suitable to be a president.

Subjects from five ethnic groups (whites, blacks, Hispanics ,Native Americans and Asians), and all ages (grouped to kids, teenagers and adults) were given profiles of ten imaginary candidates for presidency.

Two of the candidates were white, two were black, two were Hispanic, two were Native Americans and two Asians. Every pair was divided into one male and one female (candidate 1 is a white male, candidate 2 is a white female, etc…).

The subjects were asked two main questions: 1. Which one of the candidates are suitable to be the president? 2. Which one of the candidates are not suitable to be the president. For each question, they could mark one or more candidate. Therefore, for each subject, each candidate got one of three marks: Suitable to be president, not suitable to be president or no opinion at all.

I wish to test whether there is a relation between the subject’s ethnicity and his/her answers. I would also like to take into account the subject’s gender and age. It’s interesting to check if people voted to candidate with the same background, if the candidate’s gender was a factor, and if people voted for candidates of another ethnicity, for which one.

I thought to summarize the results using multiple response analysis. The relation between a the ethnicity and a certain candidate can be tested using a contingency table or modeled using a linear model, however, I wonder if there is a smarted way, or more correct way to answer this question in general, and not per candidate? Is there a model or test that handle this problem? If not, are there any data manipulations I can do, such as creating new variables that can help answering the question?

Any advice will be most appreciated. I attach a table for example so you can see how it looks like. I have more than 300 subjects. Thank you !