Kiwi Shading Data

Description

The kiwishade data frame has 48 rows and 4 columns. The data are from a designed experiment that compared different kiwifruit shading treatments. There are four vines in each plot, and four plots (one for each of four treatments: none, Aug2Dec, Dec2Feb, and Feb2May) in each of three blocks (locations: west, north, east). Each plot has the same number of vines, each block has the same number of plots, with each treatment occurring the same number of times.

Usage

kiwishade

Format

This data frame contains the following columns:

yield

Total yield (in kg)

plot

a factor with levels east.Aug2Dec, east.Dec2Feb, east.Feb2May, east.none, north.Aug2Dec, north.Dec2Feb, north.Feb2May, north.none, west.Aug2Dec, west.Dec2Feb, west.Feb2May, west.none

block

a factor indicating the location of the plot with levels east, north, west

shade

a factor representing the period for which the experimenter placed shading over the vines; with levels: none no shading, Aug2Dec August - December, Dec2Feb December - February, Feb2May February - May

Details

The northernmost plots were grouped together because they were similarly affected by shading from the sun in the north. For the remaining two blocks shelter effects, whether from the west or from the east, were thought more important.

Source

Snelgar, W.P., Manson. P.J., Martin, P.J. 1992. Influence of time of shading on flowering and yield of kiwifruit vines. Journal of Horticultural Science 67: 481-487.

References

Maindonald J H 1992. Statistical design, analysis and presentation issues. New Zealand Journal of Agricultural Research 35: 121-141.

Examples

print("Data Summary - Example 2.2.1")
attach(kiwishade)
kiwimeans <- aggregate(yield, by=list(block, shade), mean)
names(kiwimeans) <- c("block","shade","meanyield")

kiwimeans[1:4,]
pause()
 
print("Multilevel Design - Example 9.3")
kiwishade.aov <- aov(yield ~ shade+Error(block/shade),data=kiwishade)
summary(kiwishade.aov)
pause()

sapply(split(yield, shade), mean)

pause()

kiwi.table <- t(sapply(split(yield, plot), as.vector))
kiwi.means <- sapply(split(yield, plot), mean)
kiwi.means.table <- matrix(rep(kiwi.means,4), nrow=12, ncol=4)   
kiwi.summary <- data.frame(kiwi.means, kiwi.table-kiwi.means.table)   
names(kiwi.summary)<- c("Mean", "Vine 1", "Vine 2", "Vine 3", "Vine 4")
kiwi.summary
mean(kiwi.means) # the grand mean (only for balanced design)

require(nlme)
kiwishade.lme <- lme(fixed = yield ~ shade, random = ~ 1 | block/plot, 
data=kiwishade)
res <- residuals(kiwishade.lme)
hat <- fitted(kiwishade.lme) # By default fitted(kiwishade.lme, level=2)
coplot(res ~ hat | kiwishade$block, pch=16, columns=3,
  xlab= "Fitted", ylab="Residuals")
 
res <- residuals(kiwishade.lme)
hat <- fitted(kiwishade.lme, level=0) # shade effects only
unique(hat) # There are just four distinct values, one per treatment
coplot(res ~ hat | kiwishade$block, pch=16, columns=3,
  xlab="Fitted", ylab="Residuals")

n.omit <- 2
take <- rep(TRUE, 48)
take[sample(1:48,2)] <- FALSE
kiwishade.lme <- lme(yield ~ shade, data = kiwishade,
                     random = ~1 | block/plot, subset=take)
VarCorr(kiwishade.lme)[4, 1]  # Plot component of variance
VarCorr(kiwishade.lme)[4, 1]  # Vine component of variance

detach(kiwishade)