How to add predictors to intercept / slope as outcomes model

[stollbej]

Hello Forum members,

My question is theoretically very trivial but somehow I cannot realise it in MLwiN. Let´s assume I have a two-level model.
I have my constant as well as Level 1 predictors that are allowed to be random at Level 2. This results in the respective Level 2 equations of these parameters to which I would like to add further Level 2 variables. However, in MLwiN each variable first appears at Level 1. If I let this variable go random it gets its own Level 2 equation which is not my goal. If you have a solution for this I would highly appreciate it!

Thanks very much,
Jake

[joneskel]

Unlike some software (such as HLM), MLwiN has level 2 variables in the main equation. If they are really level 2 variables, they will show with a j subscript, and not an ij subscript, as they are constant at level 2. If they do not show as j there is probably something wrong with the sorting of the variable and you can check this by looking at the variable in question and the level 2 indicator variable, they should only change in harmony.

If you want to use a higher-level variable to try and explain the random slopes associated with a level -1 predictor, you first need to create a cross-level interaction between the level 1 predictor and the level 2 group variable. [ In the equations window - add term - order 1 and then add both the level 1 and level 2 variables as an interaction - the new variable will then have the ij not j subscript]

Chapter 6 of the User manual deals with this and the inclusion of the higher level variable is known as estimating 'contextual' effects

[stollbej]

Thank you very much for your swift reply and your time!

This was exactly what I was looking for and yes my basis for comparison was indeed HLM.

I wish you a nice weekend,
Jake