[R] Missing data and LME models and diagnostic plots

Peter Flom peterflomconsulting at mindspring.com
Wed Oct 21 19:13:48 CEST 2009


Running R2.9.2 on Windows XP

I am puzzled by the performance of LME in situations where there are missing data.  As I understand it, one of the strengths of this sort of model is how well it deals with missing data, yet lme requires nonmissing data.  


m1.mod1 <- lme(fixed = math_1 ~ I(year-2007.5)*TFC_,
                data = long,
                random = ~I(year-2007.5)|schoolnum)

causes an error in na.fail.default, but adding na.action = "na.omit" makes a model with no errors.  However, if I create that model, i.e.

m1.mod1 <- lme(fixed = math_1 ~ I(year-2007.5)*TFC_,
               data = long, 
               random = ~I(year-2007.5)|schoolnum,
               na.action = "na.omit")

then the diagnostic plots suggested in Pinheiro & Bates produce errors; e.g.

plot(m1.mod1, schoolnum~resid(.), abline = 0)

gives an error "could not find function "NaAct".

Searching the archives showed a similar question from 2007, but did not show any responses.

Thanks for any help



Peter L. Flom, PhD
Statistical Consultant
Website: www DOT peterflomconsulting DOT com
Writing; http://www.associatedcontent.com/user/582880/peter_flom.html
Twitter:   @peterflom

More information about the R-help mailing list