[R-sig-ME] "Multilevel Modeling Using R" and variance estimates
Douglas Bates
bates at stat.wisc.edu
Thu Oct 30 15:25:07 CET 2014
I recently received an email advertisement from CRC Press which mentioned
the recently published book "Multilevel Modeling Using R". In the errata
section of the web site for the book, http://www.mlminr.com/, the first
erratum concerns the concerns the calculation of the intraclass correlation
coefficient where the estimated standard deviations of the random effects
and residuals are used instead of the estimated variances.
The description of the erratum continues with the statements, "This error
of using the standard deviation instead of the variance continues for
several others examples. The lme4 package reports information on random
effects in terms of the standard deviations, not in terms of the variances."
I haven't looked at a copy of the book as it is not yet available at our
library. However, these statements seem peculiar to me. Does the first
sentence mean that the confusion of standard deviations and variances
continues throughout the book? If so, that seems like a rather fundamental
mistake.
Also, with regard to the lme4 package reporting standard deviations, not
variances, I thought we reported both. Is mistaking the column labelled
Std. Dev. for variance estimates a common problem?
Finally, most books with titles ending in "With R" or "Using R" provide an
R package with the data. The data in http://www.mlminr.com/data-sets are
in CSV and SPSS formats. Why SPSS?
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list