Setting up a THREE-LEVEL MODEL

[drddtorres]

Hi,

I just want to make sure I'm setting up my three-level growth model correctly. I'm familiar with time-varying variables like age and age-squared being at level 1 and time-invariant individual-level variables being at level 2; individuals are nested in level 3. Is this the same convention that MLwiN uses? How should the following variables be labeled?

1. Year (time-varying variable denoting year)
2. Mom ID
3. Child ID

Also, when ticking the ijk, is it only necessary to do that for the constant?

Diego

[joneskel]

I am guessing a bit from what you have asked, but I think you are fitting a three level model in which measurement occasion is nested within child within mother. If this is correct, then you must sort on these so that all children of the same mother are next to each other in the worksheet and similarly occasion variables for the same child are also next to each other. Remember in Mlwin you have to carry all other relevant variables when you sort. You then specify i to be the occasion index of the response, j to to be the child and k to be the mother indicator. Ticking on i, j and k for the constant gives you the random intercepts at each level. You can evaluate that all is well by looking at the hierarchy viewer once you have specified this model.

Variables at each level can then be simply added to the model and if all is well they should automatically show up with the correct subscript. Thus for example mothers education would be subscripted k.

Kelvyn Jones