Re: Ang. [Gretl-users] t-distribution in p-value finder
On Fri, 8 Jun 2007, Sven Schreiber wrote:
I would argue that a noncentral t is not t just like the noncentral
chi^2 is not chi^2. Or are there plans to subsume the noncentral chi^2
under the chi^2 as well?
Anyway I needed to explain to students in class today what it means or
why it's ok to leave at the defaults, and the small confusion that it
created is unnecessary IMHO. So if non-central t is really needed (is
it?) couldn't it be separated from t more clearly?
I've removed the offending parameters from the "t" tab in the
p-value finder.
The notion was this: if you wanted to compute a p-value for a
certain coefficient estimate, given a hypothesized value of the
parameter in question and a standard error, you could enter the
latter two numbers as the "mean" and "standard deviation"
respectively. That is, it would automate the calculation of the
actual t-value before finding the p-value. But if this is
confusing I'm OK about removing it.
Allin.