I am performing The Fisher-type test to test for stationarity for a panel data. There are 4 p-values for this test right? the inverse chi-squared (P), inverse-normal (Z), and inverse-logit (L*) and a modified version of the inverse chi-squared (Pm). The problem is that I find, for example, three of them reject the null hypothesis while one of them (Z) is not rejecting it. is this possible? and how to decide in this case? Thank u.
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
These are four different tests, thus different test statistics and different p-values, not "4 p-values for this test". However, the tests are conducted in a similar fashion, which is why I would call them Fisher-type tests or Fisher-style tests.
It is absolutely possible to get different indications by these tests.
Btw: I don't think L* is really called "inverse-logit". But I think I recall having seen a reference where it is called "inverse-logit", maybe by accident.
-
$\begingroup$ Thank u for ur answer. so if there is at least one not rejecting H0, do I assume non stationarity? $\endgroup$ Commented Feb 23, 2018 at 11:17