I don't really understand this, but it sounds highly optimistic to me. It sounds as if you want to look at the very very fine detail of the t distribution and virtually no simulation can ensure reliable results on that. It seems all too likely that mundane details such as using finite arithmetic and quite what the random number generator does (however good it is)) could easily get in your way. Others should have smarter things to say.
Nick
[email protected]
Victor M. Zammit
By database for a t table I mean having enough t values to map the
confidence intervals associated with increments of t-values.I assume that
the closer the database is closer to infinity,ie large enough,the more
convergent the mapping gets to the values in the official t-table.
----- Original Message -----
From: "Nick Cox" <[email protected]>
> That's removed one source of doubt. What remains is what you mean
precisely by a database for a t table, but go to the help for density
functions, and look under Student's t.
> Victor M. Zammit
>
> By infinite I mean a large enough number of samples of thirty each,that
> brings convergence to my results..The litmus test would be,convergence to
> the official t-table.
> If there is some analytic solution,that would be great too,but I do not
know
> where I could getthat from.
>
> Nick Cox
>
> > What do you mean by "infinite" precisely? There are various mathematical
> meanings, courtesy of Cantor, but I doubt that you mean any of them.
> >
> > Otherwise I can think of three answers to your question.
> >
> > 1. -help limits- tells you the limits your Stata has in terms of what it
> can do; your machine and OS probably cannot oblige.
> >
> > 2. You can set up a dataset with 30 observations and keep sending
results
> from a loop to a file outside Stata, but in terms of doing anything with
> those results in Stata you are back to answer 1.
> >
> > 3. What you are doing probably has an analytic solution so that
> computation is unnecessary. In particular, Stata's functions will surely
> print out a t-table better than any you can get by simulation.
>
> Victor M. Zammit
>
> > I need to t-test an infinit number of random samples of size 30 ,from an
> > infinite,normally distributed population.Each t-value is saved and then
> > appended together to form a database for t table.The problem is that I
get
> > constrained by memory regardless of the size of the memory in my Stata.I
> > have Version 9 .Is there a way of getting around this constrained.?
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
