At 10:13 AM 5/11/99 +0000, you wrote:
>The theoretical distribution is most definitely not an F-distribution,
>though the shape may be reminiscent.
If we knew what the distribution is theoretically, we could fit it. I'm a
former temporary part-time adjunct instructor of statistics for a minor
university (really!), but it has been a long time and I don't have most of my
books at hand.
>Since the pdf of the normal distribution does not have a "pretty"
>analytic integral, the function I[0,x] is complex.
I've got a decent numerical integration routine (that I use for integrating the
Gaussian dist.)
>All assuming that the underlying distribution of the source data is
>normal - or near enough to make no practical differerence!
>
If you're talking about the actual data George gave us - it is far from
normal. At first I was going to do a Kolmogorov-Smirnov test to see if it was
close to a normal, but that is really designed for the raw data. Since the
data was in a histogram I used the Chi-squared test. The data posted was far
from a normal distribution (even just the right half of it was far from a
normal).
+------------------------------------------+
| Jud McCranie [EMAIL PROTECTED] |
+------------------------------------------+
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
