I would like to have your suggestions regarding the kind of statistical analysis that would fit the nature of my data.

Participants have seen 12 videos, in each of which an emotion was conveyed and they had to guess which emotion was conveyed in each emotion (I plan to do a separate analysis for each video, so 12 similar analyses in total). Participants were presented with 36 words and had to choose the word that correspond to the emotion conveyed. The independent variables will be categorical (e.g. native speakers versus learners vs. non-speakers of Chinese).

I would like to not focus on whether they chose the ”right” emotion which was intended to be conveyed in the video (so it’s not a dichotomous dependent variable, I think), but I would like to see the different patterns between my groups: e.g. what are the first, second, third, fourth,… most chosen words among the native speakers vs. among the learners vs. among the non speakers, and are the differences significant?

Basically, I would like to see where the differences in this bar chart are significant (this corresponds to the data for the video in which the emotion "afraid" was conveyed, broken down by Status of Chinese enter image description here


1 Answer 1


I think that question could be best answered using a multinomial logit model. That will allow you to estimate how the probability of choosing a certain word over another changes with characteristics of the person.

A more simple alternative would be a Chi-square test of association which will basically tell you whether there is a different pattern in the words chosen between different groups. The down-side of that test is that you can only do the test one characteristic at the time, e.g. word choice differences between native and non-native, then word choice differences between men and women etc.

  • $\begingroup$ Thank you Maarten. I looked into multinomial regression. If I'm right, it will compare one category of the DV with all the other categories of the DV (at least that is what you get in the SPSS output), in my different groups (IVs). Do you know if there is a way of comparing all the categories of the DV with each other ? $\endgroup$
    – Pernelle
    Jul 22, 2018 at 14:21
  • $\begingroup$ @Pernelle Yes that is right. Based on your bar graph above it would for example report how the probability of choosing neutral versus tense changes if you are native Chinese, versus a foreign language person, and the same for all other comparisons between neutral and the other categories. I'm not sure if there is a way of comparing all categories, although you could of course vary the reference level. $\endgroup$ Jul 23, 2018 at 14:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.