>>hypothesis that people are more likely to watch movies with multiple superhero characters<<
If you're testing for "more likely" then what is survey data going to tell you?
Because for that hypothesis, all you need are some statistics:
(sales) for [five-to-ten different single superhero movies], versus
(sales) for [five-to-ten different multiple superhero movies],
where (sales) = theater tickets + video purchases + downloads
That's really about as close to an apples-to-apples comparison as can be done. You'd probably want to select the movies based on an arbitrary rule, like "ten most recent" or whatever, and you would want a big enough sample size to account for the fact that some superhero movies are way way more popular than others.