CODEWITHSUNDEEP
r
macintosh
macintosh

Reputation: 21

Wilcoxon test with chicken's weight

data(ChickWeight)
head(ChickWeight)
plot( ChickWeight$Time, ChickWeight$weight, col=ChickWeight$Diet)
chick = reshape(ChickWeight, idvar=c("Chick","Diet"), timevar="Time",
                direction="wide")
head(chick)
chick = na.omit(chick)

Perform a t-test of x and y, after adding a single chick of weight 200 grams to x (the diet 1 chicks). What is the p-value from this test? The p-value of a test is available with the following code: t.test(x,y)$p.value

Do the same for the Wilcoxon test. The Wilcoxon test is robust to the outlier. In addition, it has less assumptions that the t-test on the distribution of the underlying data.

When I try do Wilcoxon test:

wilcox.test(c(x, 200), y)

I get this error:

Warning message:
In wilcox.test.default(c(x, 200), y) :
  cannot compute exact p-value with ties

Upvotes: 1

Views: 119

Answers (1)

TarJae
TarJae

Reputation: 79194

Use exactRankTests::wilcox.exact:

If x is a weight for example time 0 e.g. chick$weight.0 and y is a weight for example time 2 e.g. chick$weight.2

Then you could do it this way:

With wilcox.test you will get a warning message:

> wilcox.test(c(chick$weight.0, 200), chick$weight.2)$p.value
[1] 6.660003e-14
Warning message:
In wilcox.test.default(c(chick$weight.0, 200), chick$weight.2) :
  cannot compute exact p-value with ties

Use exactRankTests::wilcox.exact() that can handle ties:

t.test(chick$weight.0,chick$weight.2)$p.value
6.660003e-14

exactRankTests::wilcox.exact(c(chick$weight.0, 200), chick$weight.2)$p.value
[1] 5.889809e-18

Upvotes: 1

Related Questions