Re: Hypothesis testing with proportions

In article <42e7512f$0$16686$9b4e6d93@xxxxxxxxxxxxxxxxxxxxxxxxxx>, <bronzing@xxxxxx> wrote:

> Just to clarify my problem (coz I'm not sure if you got me right and
> the proposed procedures are applicable): We made 45 gels from tissue
> samples of one patient which contain partially mutant partially
> wild-type DNA. From the IODs (integrated optical desities) of the
> bands on the gels we calculated for each of the 45 gels the
> proportion of mutant DNA contained in the gel. The mean value of all
> the 45 proportions is 0.291 with a standard deviation of 0.035. The
> proposed web-site deals with samples taken from populations, each
> sample producing an integer number. (Am I wrong?) I don't know how
> this is comparible to my problem. I posted the same question to the
> forim of OriginLab (the software I'm using for my
> calculations). Here's the answer I got until now. What do you think?

You should just do a t test. Your proportions are not integer counts
out of some integer total, so any method designed for that situation
is inapplicable. In theory, the t test assumes that the values are
normally distributed, but it's robust to departures from this
assumption, especially if the sample size isn't too small (45 is big
enough), and the departures from normalily don't take the form of
occassional extreme values. Since your values are between 0 and 1,
there can't be any really extreme values.

The more crucial assumption is that the values are independent.
There's no obvious reason to think this assumption is violated from
your description, but it's something to keep in mind.

Ignore the advice from "Reef Fish", who is a raving idiot.

----------------------------------------------------------------------------
Radford M. Neal radford@xxxxxxxxxxxxxx
Dept. of Statistics and Dept. of Computer Science radford@xxxxxxxxxxxxxxxxxx
University of Toronto http://www.cs.utoronto.ca/~radford
----------------------------------------------------------------------------
.