The wikipedia page for multilevel model states:

The dependent variable must be examined at the lowest level of analysis.

I am interested in predicting measures at higher levels of analysis. In simple cases, I understand there are ways to aggregate lower level data. However, if we had unequal number of lower level observations, we should be more confident in some of the means, but this would not be taken into account if we just used the aggregate measures at the higher level. I believe a Bayesian model could work here. In a neural network, could a LSTM wrap up lower level data, and then send it all to predict higher level data? Are the other approaches or resources for this kind of task?



Your Answer

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