One tailed or Two tailed only, question, something, different

soverby

This is sort of a newb question, but when they are asking us to test if X is significantly different from Y blah blah we are supposed to use two tailed, and if they ask us to do a confidence interval like X +- (t value)(Standard error) , we use two tailed also.

We use one tail for F stat only?

I know you use one tail when you are testing if something is greater than or less than, and two tailed to test if it's equal to something.

zhaoyp

I'd like to ask 1 also. I understand that for 2-tailed test, the rejection regions are at the 2 tails. For one-tailed t-test, let's say null is b1<=5, then where is the rejection region, in the positive side or negative side? Thanks.

RMontgomery

freakingout...

here you would have

H0: b1 <= 5
Ha: b1 > 5

you would accept it if b1 <= 5... right.

so your rejection region is to the right... Look at the Alt to determine your rejection / acceptance region.

Two tailed:
H0: b1 = 5
H1: B1 <> 5

same thing => at the extremes are the rejection region.

CP