Diet: Diet Effect on Dry Matter Intake

In byandell/pda: Practial Data Analysis for Designed Experiments

Description

This experiment (Wattiaux, Combs and Shaver 1994) involved 60 cows initially, although some were lost during the study. The cows were randomly assigned to one of 6 diets and followed for a number of weeks. Diets were begun after the third week, allowing the animals some initial time to adjust to their new environment. Interest focuses on the effect of diet on the average dry matter intake (dmi), the amount of food eaten by each cow. The data you have are a baseline value (covar = dmi for week 3), average dmi during subsequent weeks, the number of subsequent weeks. In addition there are cow and block identifiers.

The experiment was a randomized block design. That is, cows were blocked by time, the first 6 cows were randomly assigned among the 6 diets, and so on. “Proper” analysis should take account of this, along with the initial capacity of each animal (its covariate of dry matter intake at 3 weeks) and possibly weighting by the number of weeks each cow was on trial.

Usage

data(Diet)

Format

Diet data frame with 12 observations on 7 variables.

[,1]	trt	factor	treatment identifier
[,2]	bwt	numeric	body weight
[,3]	bwg	numeric	body weight gain
[,4]	fi	numeric	diet intake
[,5]	fe	numeric	diet(?)
[,6]	cla	numeric	CLA amount
[,6]	la	numeric	LA amount

Source

Michel Wattiaux, International Agriculture Program, CALS, UW-Madison

References

Wattiaux MA, Combs DK and Shaver RD (1994) “Lactational responses to ruminally undegradable protein by dairy cows fed diets based on alfalfa silage”, J Dairy Science 77, 1604-1617.

Examples

data( Diet )

Diet$trt <- factor(Diet$trt)
Diet$block <- factor(Diet$block)
#
# ordinary anova
Diet.fit <- aov(dmi ~ trt, Diet)

# ancova with covar = first 3 weeks dmi
Diet.cov <- lm(dmi ~ covar + trt, Diet, na.action=na.omit)

# ancova with blocking and weighting by number of weeks on trial
Diet.blk <- lm(dmi ~ covar + block + trt, Diet,
   weights=Diet$weeks, na.action=na.omit)

# Figure F17.2: Diet Adjusted for Initial Capacity
## regression lines in following plot are not parallel
Diet = Diet[!is.na(Diet$covar), ]
Diet$pred = predict(Diet.cov)
trellis.par.set(theme=col.whitebg())
print(xyplot(dmi ~ covar, Diet, groups = trt, type = "p",
  pch = levels(Diet$trt), cex = 1.5,
  col = 1 + as.numeric(levels(Diet$trt)),
  panel = function(x,y,...) {
    panel.superpose(x,y,...)
    pred = predict(Diet.cov)
    for(i in levels(Diet$trt)) {
      ii = Diet$trt == i
      i = as.numeric(i)
      panel.lines(Diet$covar[ii], Diet$pred[ii],
        lty = i, col = i + 1)
    }
    ## SD bar
    lsd.bar(Diet.cov, xpos = 24, ypos=20, cap = "SD")
  },
  xlab = "covariate = first 3 weeks",
  ylab = "dry matter intake (dmi)",
  main = "Figure F:17.2 Diet Adjusted for Initial Capacity"))
data.frame(se.bar(26, 20, std.dev(Diet.cov), cap = "SD"))

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