interpretation of topic

[Teresa]

Hi,

sorry I'm new in the world of meta analysis and I've one simple question: How should I interpret the topics? for example, each topic would correspond to the first word in the list?

Thank you!!

[Taylor Salo]

Topic models define the topics in terms of probability distributions across terms. The topics on Neurosynth show the top 30 or so terms, but the first term (while having the highest probability) is not a good summary of the topic on its own. The best approach is probably to take all of the terms into consideration, and summarize those terms according to whatever common theme there might be. Of course, latent Dirichlet allocation (the topic model used in Neurosynth) is an unsupervised approach, so there's no guarantee that topics will map cleanly onto meaningful categories, but that's something you'll need to take into account. I hope that helps.