R help - Apr 2012 - Multivariate Multilevel Model: is R the right software for this problem

Hello,

I've been trying to answer a problem I have had for some months now and
came across multivariate multilevel modeling. I know MPLUS and SPSS quite
well but these programs could not solve this specific difficulty.

My problem:
9 correlated dependent variables (medical symptoms; categorical, 0-3), 5
measurement points, 10 time-varying covariates (life events; dichotomous,
0-1), N ~ 900. Up to 35% missing values on some variables, especially at
later measurement points.

My exploratory question is whether there is an interaction effect between
life events and symptoms - and if so, what the effect is exactly. E.g. life
event 1 could lead to more symptoms A B D whereas life event 2 could lead
to more symptoms A C D and less symptoms E.

My question is: would MMM in R be a viable option for this? If so, could
you recommend literature?

Thank you
--T

	[[alternative HTML version deleted]]

I recommend looking at chapter 6 of Paul Allison's Fixed Effects Regression
Models.  This chapter outlines how you can use a structural equation modeling
framework to estimate a multi-level model (a random effects model).  This
approach is slower than just using MLM software like lmer() in the lme4 package,
but has the advantage of being able to specify correlations between errors
across time, the ability to control for time-invariant effects of time-invariant
variables, and allows you to use the missing data maximum likelihood that comes
in structural equation modeling packages.

Andrew Miles
Department of Sociology
Duke University

On Apr 6, 2012, at 9:48 AM, Eiko Fried wrote:
> Hello,
> 
> I've been trying to answer a problem I have had for some months now and
> came across multivariate multilevel modeling. I know MPLUS and SPSS quite
> well but these programs could not solve this specific difficulty.
> 
> My problem:
> 9 correlated dependent variables (medical symptoms; categorical, 0-3), 5
> measurement points, 10 time-varying covariates (life events; dichotomous,
> 0-1), N ~ 900. Up to 35% missing values on some variables, especially at
> later measurement points.
> 
> My exploratory question is whether there is an interaction effect between
> life events and symptoms - and if so, what the effect is exactly. E.g. life
> event 1 could lead to more symptoms A B D whereas life event 2 could lead
> to more symptoms A C D and less symptoms E.
> 
> My question is: would MMM in R be a viable option for this? If so, could
> you recommend literature?
> 
> Thank you
> --T
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

	[[alternative HTML version deleted]]

I recommend looking at chapter 6 of Paul Allison's *Fixed Effects
Regression Models*.  This chapter outlines how you can use a structural
equation modeling framework to estimate a multi-level model (a random
effects model).  This approach is slower than just using MLM software like
lmer() in the lme4 package, but has the advantage of being able to specify
correlations between errors across time, the ability to control for
time-invariant effects of time-invariant variables, and allows you to use
the missing data maximum likelihood that comes in structural equation
modeling packages.

Andrew Miles

	[[alternative HTML version deleted]]