Post hoc tests after ANOVA

In the comparison of all groups versus 1 control: What if we want cars with 5 gears as control?

Which comparison will be done by the third row of the contrast matrix?

If the spreads of the groups are not equal, can you still use Tukey’s test?

When the spreads of the groups are not equal and you want to do specific comparisons, the best thing to do is pairwise t-tests with Welch correction:

> p1 <- t.test(mtcars[mtcars$gear==3,”mpg”],mtcars[mtcars$gear==4,”mpg”])$p.value
> p2 <- t.test(mtcars[mtcars$gear==3,”mpg”],mtcars[mtcars$gear==5,”mpg”])$p.value

However, you still need to correct for doing multiple (2) tests on the same data set. How to do this correction?