fat: Study of influence of menarche on changes in body fat...
In ALA: Data sets and examples for the book "Applied Longitudinal Analysis"

Description Usage Format Details Note Source References Examples

The data are from a prospective study on body fat accretion in a cohort of 162 girls from the MIT Growth and Development Study. The study examined changes in percent body fat before and after menarche. The data represent a subset of the study materials and should not be used to draw substantive conclusions.

data(fat)

A data frame with 1049 observations on the following 5 variables.

id: a factor with 162 levels
age: a numeric vector
age.menarche: a numeric vector; age at menarche
time.menarche: a numeric vector; time since menarche
percent.fat: a numeric vector

At the start of the study, all of the girls were pre-menarcheal and non-obese, as determined by a triceps skinfold thickness less than the 85th percentile. All girls were followed over time according to a schedule of annual measurements until four years after menarche. The final measurement was scheduled on the fourth anniversary of their reported date of menarche. At each examination, a measure of body fatness was obtained based on bioelectric impedance analysis and a measure of percent body fat (%BF) was derived. In this data set there are a total of 1049 individual percent body fat measurements, with an average of 6.4 measurements per subject. The numbers of measurements per subject pre- and post-menarche are approximately equal.

Original variable names have been adapted to R conventions.

http://biosun1.harvard.edu/~fitzmaur/ala

Phillips SM, Bandini LG, Compton DV, Naumova EN, Must A (2003) A longitudinal comparison of body composition by total body water and bioelectrical impedance in adolescent girls. Journal of Nutrition 133:1419-1425

str(fat)
summary(fat)

if (require(lattice)) {
    ## Fig. 8.5 (roughly)
    xyplot(percent.fat ~ time.menarche, data=fat, groups=id, type="b",
           cex=0.5, col=1,
           xlab="Time relative to menarche (weeks)",
           ylab="Percent body fat")
    ## Fig. 8.6 (roughly)
    xyplot(percent.fat ~ time.menarche, data=fat,
           cex=0.5, col=1,
           xlab="Time relative to menarche (years)",
           ylab="Percent body fat",
           panel=function(x, y, ...) {
               panel.abline(v=0, lty=2)
               panel.xyplot(x, y, ...)
               panel.loess(x, y, ...)
           })
}

if (require(lme4)) {
    fatNew <- within(fat, {
        ## Create the stage factor -- this is what is needed to make the
        ## same interpretations as in the book
        stage <- cut(time.menarche,
                     breaks=c(floor(min(time.menarche)), 0,
                       ceiling(max(time.menarche))),
                     labels=c("pre", "post"))
        ## But this is what is actually used
        stage.tij <- pmax(time.menarche, 0)
    })
    summary(fatNew)
    ## Model in p. 218
    (fm1 <- lmer(percent.fat ~ time.menarche + stage.tij +
                 (time.menarche + stage.tij | id), data=fatNew))
    ## which is the same as a model using the interaction with the stage
    ## factor; i.e. no interest in intercept differences between stages,
    ## only in slope differences
    (fm1b <- lmer(percent.fat ~ time.menarche + time.menarche:stage +
                  (time.menarche:stage | id), data=fatNew))
    ## Table 8.7
    VarCorr(fm1)[[1]]
    ## Fig. 8.7 (roughly)
    set.seed(1234); rndID <- sample(levels(fatNew$id), 2)
    tm <- with(fatNew, seq(floor(min(time.menarche)),
                           ceiling(max(time.menarche))))
    fitted.pf <- fitted(fm1)
    avg.modmat <- cbind(1, tm, pmax(tm, 0))
    pred.fixef <- avg.modmat %*% fixef(fm1)
    plot(pred.fixef ~ avg.modmat[, 2], type="l", ylim=c(5, 35), lwd=2,
         xlab="Time relative to menarche (years)",
         ylab="Percent body fat")
    with(fatNew, {
        points(time.menarche[id == rndID[1]], percent.fat[id == rndID[1]])
        lines(time.menarche[id == rndID[1]], fitted.pf[id == rndID[1]])
        points(time.menarche[id == rndID[2]], percent.fat[id == rndID[2]], pch=2)
        lines(time.menarche[id == rndID[2]], fitted.pf[id == rndID[2]], lty=2)
    })
}