I am confused which tests do i use a two tailed test and which do i use one tailed test at a certain level of significance, let’s say level of significance of 5%?
The wording of the question will tell you. If you see “greater than” or “less than”, that’s one-sided; if you see “not equal”, that’s two-sided.
Eg. you want consider that deviations in some process exist which can cause the losses (a negative deviation) or the extra gains (a positive deviation). You suspect that deviation exists but want to be sure.
If you simply want to prove that deviation exists, you will take a two tailed test with H0 = 0 = no deviation or H1 not equal to 0 = there is a deviation.
If a positive deviation is something not to worry about, you’ll concentrate on testing only if a negative deviation exists, thus you’ll take a one tailed test with H0 <= 0 no negative deviation or H1 > 0 = there is a negative deviation which might led to significant losses.
You’ll take a test at a various significance level, 10 %, 5 % or 1 % depending on impact of deviation to your process. Here you must take a care if one or two tailed test is taken.
Thanks this is really beneficial!