[R] Joint Test for Means of different samples (panel)?

Andreas Klein klein82517 at yahoo.de
Fri Nov 6 12:38:31 CET 2009


Hello everybody.

Does anybody know a statistical test for the following problem?


I have 5 samples (time series).
Each consist of 100 dependent observations.
The samples are independent of each other.
It is similar to a panel of five cross-sections and 100 time points.


I have two different test problems:

1. Problem:
"sample1 = sample2 = sample4 = sample5 = 0 and sample3 != 0"

2. Problem:
"sample1 >= 0, sample2 >=0, sample4 <=0, sample5 <= 0 and sample3 != 0"


Does anybody know a test, that would handle the null-hypothesises of problem 1 and 2? Or does anybody know how to solve my problems?

I thought that sth. like a Wald-Test would work, but I didn't get the point of transferring it to my problems, because I only know the Wald test from joint hypothesises from regressions.
And especially problem 2 has a mixture of a one sided (<=) and a two sides (!=) test problematic.


I hope someone can help me.


Sincerely
Andres Klein.







More information about the R-help mailing list