[R] Producing residual plots by time for lme object

[Deepayan Sarkar]

On Fri, Jul 2, 2010 at 11:40 AM, Gregory Gentlemen
<gregory_gentlemen at yahoo.ca> wrote:
> Fellow R-users,
>
> I have a longitudinal data set with missing values in it. I would like to produce a residual plot for each time using panel.xyplot function but I get an error message. Here's a simple example,
>
> library(nlme)
> set.seed(1544)
> longdata <- data.frame(ID=gl(10,1,50), y=rnorm(50), time = as.numeric(gl(5,10,50)), x = rnorm(50))
> longdata$y[5] <- NA
> longdata$x[35] <- NA
>
> modlme<- lme(fixed=y ~ time + x,
>                 random= ~ 1 | ID,
>                 na.action=na.exclude, data=longdata)
>
>
> plot( modlme, abs( resid(., type = 'p')) ~ fitted(.) | time,
>           panel = function(x, y, ...) {
>                 panel.xyplot( x, y, ...)
>                 panel.loess( x, y,...)
>           })
>
> where the last call produces the error message
> Error in `[[<-.data.frame`(`*tmp*`, j, value = c(1, 1, 1, 1, 1, 1, 1,  :
>   replacement has 48 rows, data has 50
>
>
> How do I fix this?

NA-handling is not working right, it seems. Would it make any real
difference to your analysis to do

modlme<- lme(fixed=y ~ time + x,
                random= ~ 1 | ID, data = na.omit(longdata))

(in which case the subsequent plot call works)?

-Deepayan