[R] question on xyplot

[Deepayan Sarkar]

On 10/3/08, hadley wickham <h.wickham at gmail.com> wrote:
> On Fri, Oct 3, 2008 at 8:04 AM, eugen pircalabelu
>  <eugen_pircalabelu at yahoo.com> wrote:
>  > Hi List,
>  >
>  >
>  > I have the following problem: I am using the multilevel package and make.univ function for available in the package and then xyplot from lattice and I want to know how could I be able to use the "coefficient" for the straight line that passes the data ?
>  >
>  > Example from help:
>  >
>  > library(multilevel)
>  > data(univbct)  #a data set already in univariate or stacked form for job satisfaction
>  > TEMP<-univbct[3*1:495,c(22,1:17)]  #converting it back to multivariate form
>  > TEMP2<-make.univ(x=TEMP,dvs=TEMP[,c(10,13,16)])  #transforming it to univariate form again
>  >
>  > xyplot(MULTDV~TIME|as.factor(SUBNUM),data=TEMP2,type=c("p","r","g"),col="blue",col.line="black",xlab="Time",ylab="SAT") # taken from Bliese Paul – Multilevel Modeling in R
>  >
>  > Now I want to be able to identify those SUBNUM that have a downwards trend like (for eg SUBNUM 7) without picking them by inspecting the plot. Is there some way how I could access these coefficients for this apparently regression line?
>
>
> You'll need to fit the models yourself, and then inspect the
>  coefficients of the results.  Here's one way using the plyr package:

Another option is 'lmList()' in the nlme (or lme4) package:

fm <- nlme::lmList(MULTDV~TIME|as.factor(SUBNUM), data=TEMP2, na.action=na.omit)
neg <- names(which(sapply(fm, coef)["TIME", ] < 0))

xyplot(MULTDV~TIME|factor(SUBNUM), data=TEMP2, type=c("p","r","g"),
       col.line="black",subset = (factor(SUBNUM) %in% neg))

-Deepayan

> [...]