问题
I was wondering if there is a way in R to do an ANOVA with unequal variances?
Imagine the following example:
x <- c(10,11,15,8,16,12,20)
y <- c(10,14,18,25,28,30,35)
d <- c(x,y)
f <- as.factor(c(rep("a",7), rep("b",7)))
# Unequal variance:
t.test(x,y)$p.value
t.test(d~f)$p.value
# Equal variance:
t.test(x,y, var.equal=TRUE)$p.value
t.test(d~f, var.equal=TRUE)$p.value
anova(lm(d~f))[[5]]
summary(aov(lm(d~f)))[[1]][5]
summary(lm(d~f))[[4]][8]
As you can see from this example the different ways of performing an ANOVA in R, in case of two groups only, always result in a p-value identical to the one obtained by a t.test with equal variances. Again, is there a way to perform an ANOVA with unequal variances?
回答1:
For this case there is oneway.test()
R> oneway.test(d~f)
One-way analysis of means (not assuming equal variances)
data: d and f
F = 6.631, num df = 1.000, denom df = 8.339, p-value = 0.03179
来源：https://stackoverflow.com/questions/11816948/parametric-anova-with-unequal-variances