I am a master's student doing some psychological research. I want to use child self-esteem as my outcome variable in an HLM model. At level one my predictors are characteristics of how a child has come to understand an important, sad, and a happy event such as how meaningful the child finds the event, or how much vulnerability the child displayed when relating the event. Then at level 2 I have quasi IV's child age and child gender.

So it is a bit contrived sounding I suspect but I have characteristics of the event-memory (the way particular children understand particular events, within child) nested in higher-order unchanging child-level variables (age and gender, between children) and I want to predict child self-esteem from those.

Other than feeling a little contrived I have come to worry about child self-esteem, my "dv", only varying at one level of the model (level 2 but not level one) the between child level. Does this sink the ship on using HLM?

Interestingly I have set it up in SPSS and actually gotten the models to converge using HLM 6.06. And I found some very interesting and theoretically meaningful coefficients. But I am worried that I am missing something or committing an error.

Can anyone help out with this admittedly diffuse and weird situation? Am I violating an HLM rule that I don't know about? Thanks so much for your help.