[R-sig-ME] visualizing effects of fixed factors and partial/whole model statistics

[John Fox]

Dear Tim,

You might take a look at the effects package, which handles objects produced
by lmer(), and generalizes the idea of (the horribly named) "least-squares
means." The functions to use directly are allEffects(), effect(), and
Effect() -- see ?effect.

I hope this helps,
 John

-----------------------------------------------
John Fox
Senator McMaster Professor of Social Statistics
Department of Sociology
McMaster University
Hamilton, Ontario, Canada




> -----Original Message-----
> From: r-sig-mixed-models-bounces at r-project.org [mailto:r-sig-mixed-
> models-bounces at r-project.org] On Behalf Of Timothy Farkas
> Sent: Sunday, October 07, 2012 10:13 AM
> To: r-sig-mixed-models at r-project.org
> Subject: [R-sig-ME] visualizing effects of fixed factors and
> partial/whole model statistics
> 
> Hi All,
> 
> I have two related practical questions concerning GLMMs. I hope my
> questions aren't too simplistic for this list, but other forums have
> been
> not been responsive.
> 
> My model is specified as follows:
> 
> lmer(countData ~ 3LevelFactor + continuous + (1|5LevelFactor),
> family=poisson)
> 
> and there is some redundancy between the fixed components and between
> the
> continuous variable and random factor, but not loads.
> 
> 1) How can I visualize the effect of the fixed factor while accounting
> for
> the continuous variable and random factor?
> 
> I'm looking to make a column graph with standard error bars that would
> be
> the multiple-regression equivalent of a partial plot. I think some
> people
> would refer to this as a "least-squared means" plot? Ideally, I could
> obtain a vector of adjusted values for the dependent variable and plot
> them
> over levels of the factor. Because redundancy is low, I'm OK with
> plotting
> raw data, but on principle I like my visualizations to accurately
> reflect
> my analysis if possible.
> 
> 2) Are there statistics like the R-sq of linear regression that will
> denote
> how much variation in my dependent variable is explained by my model?
> What
> about statistics like partial-R-sq, that denote the amount of variation
> explained by individual fixed parameters?
> 
> Cheers,
> 
> tim
> --
> 
> Timothy E. Farkas
> Dept. Animal and Plant Sciences
> Alfred Denny Building
> University of Sheffield
> Western Bank
> Sheffield S10 2TN
> United Kingdom
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models