Your problem seems to be more focused on
model specification and interpretation. If you want to see whether grouping
variable is required or not you can use “anova” for example to
compare a model with no grouping variable versus the multi-level model.
The output from the multilevel model
can also indicate the relationship between units and clusters. For example how the
magnitude of the estimated variance for a random effect in compares with a
fixed effect indicates whether
a) The amount of the variability in the fixed effect(or rate of change
in the response due to specified variable) from group to group.
b) A grouping variable is necessary
I hope this helps
Stella
From:
s-news-owner@lists.biostat.wustl.edu on behalf of Joyce Kwan
Sent: Wed 4/9/2008 5:23 AM
To: s-news@lists.biostat.wustl.edu
Subject: [S] Is s-plus able to
study the effect of individual-level variable on group outcome
I would like to ask if it's possible for me to fit a multilevel model
in S-Plus that enable me to study the effect of individual-level variable on
group-level variable. For example, whether individual's goal orientation will
affect teams' performance. So far as I see, most of the analysis
regarding multilevel modelling are studying how group-level variable
affect individual's repsonse. I'm wondering if it's possible for us to do the
other way round in s-plus.