I am comparing the difference in mean topic proportion (theta / document-topic-probability) of two groups. How can I interpret whether the differences are small or large, as topic proportions depend on the number of topics k?

For example:

Topic 1: group difference = 1.01 % Topic 2: group difference = 0.79 % Topic 3: group difference = 0.10 %

The LDA-based model comprised k = 100 topics. So the average topic proportion is 1/k = 1 %, which seems to be a necessary starting point for interpretation.

Based on that, can any meaningful cut-off be made to differentiate between small and large difference?

Any ideas are appreciated.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.