How to determine if the results to a multiple choice question was statistically significant I did a survey and had 41 people answer a multiple choice question, where they could only pick one answer. The frequencies included:

*

*$A = 25$

*$B = 12$

*$C = 4$

*$n = 41$
I want to know if there is a statistically significant difference between results. After reading it seems like doing a chi-square test with post-hoc testing would be best.
However, I am struggling to operationalize this / create my chi-square cross-tab. It over estimates my sample size.
Is the issue because they all came from one question, so they aren't really independent? Should I be attempting to compare these proportions in another way? Or should I not being doing it at all?

 A: Remember there are different types of chi-squared tests, so the answer to this question, as highlighted by Dave and Whuber, lies in the question you are trying to answer. Based on what you said, it's difficult to know exactly what you are looking for, so based off my intuition this is what I can see.
If you are trying to test whether or not the choice between A, B, C, or D shares the same frequency distribution, you could consider a chi-squared goodness of fit test. Here you would be assessing the following question: are there equal proportions of answers between A, B, C, or D? Perhaps you believe some people are far more likely to answer A or far less likely to answer C. You could at least use the chi-squared test as an omnibus test of this question.
What your "yes" or "no" factor is here is unclear, but it looks like you have attempted to form a contingency table for a chi-squared test of independence. I'm going to take a guess that this isn't what you are trying to do. The main purpose of this form of chi-squared test is to try to see if two factors are unrelated. Unless there is some other factor you are trying to address ("did this person cheat on the test: yes/no"), it is unlikely this would answer that question. You can certainly correct me if I am wrong with my assumption, but your table doesn't seem to highlight any clear factor in that regard.
For post hoc tests, you can use the residuals from either of these chi-squared tests. Since you are using SPSS, there is a three part series here you can look at. There are other chi-squared tests that are probably not what you are looking for, but an overview can be found here.
