Hardy: Cold Hardiness Adaptation

In byandell/pda: Practial Data Analysis for Designed Experiments

Description

Scientists in horticulture have been interested in how plants adapt to cold climates such as we find in Wisconsin. They have noticed that plants which have been previously ‘acclimatized’ suffer less damage. Palta and Weiss (1993) conducted a controlled laboratory experiment to examine two species of potatoes (1, 2) subjecting plants to one of two acclimatization regimes (1 kept in cold room; 0 kept at room temperature) for several days. Plants were later subjected to one of two cold temperature treatments (1 just below freezing, -4 C; 2 way below freezing, -8 C). Two responses were measured, both being damage scores for photosynthesis and ion leakage. Initially there were 80 plants, but some were lost during the experiment. Your analysis should take care to address the imbalance due to unequal sample sizes for the different factor combinations.

Usage

data(Hardy)

Format

Hardy data frame with 75 observations on 6 variables.

[,1]	potato	factor	potato identifier
[,2]	regime	factor	hardiness regime
[,3]	temp	factor	temperature
[,4]	code	factor	plot code
[,5]	photo	numeric	photosynthesis score
[,6]	leak	numeric	ion leakage score

Source

Jiwan P Palta (mailto:[e-mail]) \& Laurie S Weiss, U WI Horticulture, (http://www.hort.wisc.edu)

References

Palta JP and Weiss LS (1993) 'Ice formation and freezing injury: an overview on the survival mechanisms and molecular aspects of injury and cold acclimation in herbaceous plants', in Advances in Plant Cold Hardiness, ed. by PH Li and L Christersson. CRC Press, Boca Raton, LA.

Examples

data( Hardy )

Hardy$potato <- factor(Hardy$potato)
Hardy$regime <- factor(Hardy$regime)
Hardy$temp <- factor(Hardy$temp)

# full model: you may want to consider various reduced models */
Hardy.fit <- aov(leak ~ potato * regime * temp, Hardy)
# Type I sums of squares
summary( Hardy.fit )
# Type III sums of squares
drop1( Hardy.fit, formula( Hardy.fit ), test = "F" )

# Least Squares means
Hardy.lsm <- lsmean(Hardy.fit)
# default plots for lm object
plot(Hardy.fit)

# diagnostic plot of resid vs predict with 0 line and +/- 1 SD
Hardy$code = factor( Hardy$code )
xyplot( resid(Hardy.fit) ~ jitter(predict(Hardy.fit)),
  groups = Hardy$code, pch = levels( Hardy$code ), cex = 1.5,
  panel = function(x,y,sd=std.dev(Hardy.fit),...){
    panel.xyplot(x,y,...)
    panel.abline(0,0,lty=2)
    panel.abline(sd,0,lty=3)
    panel.abline(-sd,0,lty=3)
  } )

# interaction plot with LSD bars (but play with this!)
lsd.plot(Hardy.fit,Hardy,c("potato","regime"),
  more = TRUE, split = c(1,1,2,1) )
# the following give a slightly different plot -- why?
attach(Hardy)
lsd.plot(potato,leak,regime, split = c(2,1,2,1))
detach()

# side-by-side interaction plots for the subsets of t=1,2
lsd.plot(Hardy.fit,Hardy,c("potato","regime"),
  Hardy.lsm[Hardy.lsm$temp==1,],
  more = TRUE, split = c(1,1,2,1) )
lsd.plot(Hardy.fit,Hardy,c("potato","regime"),
  Hardy.lsm[Hardy.lsm$temp==2,],
  split = c(2,1,2,1) )

byandell/pda documentation built on May 13, 2019, 9:27 a.m.