Re: Is this sample representative of the population?

In article <43a66ace$0$4338$626a54ce@xxxxxxxxxxxx>,
"Z" <zingerNOSPAM@xxxxxxx> wrote:

> "David Winsemius" <dwin$emiu$@comcast.net> a écrit dans le message de
> news:dwin$emiu$-A6EA19.01435718122005@xxxxxxxxxxxxxxxxxxxxxxxxxxx
> > Fine. You are worried about the second numbers in each row being a
> > subset of the first? Subtract the second from the first, then do a
> > general test of association on the proportions of responders and
> > non-responders. You don't even need a statistical package. There are
> > online resources.
> >
> > You could be underestimating your client. It seems perfectly possible
> > that the client may be testing your ability to do a simple statistical
> > test. This is first year stuff, even in marketing/business/psychology
> > curriculums.
> >
>
> Thanks. Do you have a link where I could see this "general test of
> association on proportions"? I've been googling over "general test of
> association" and "proportions" and not getting anywhere.As to the textbooks
> I have, they're all in French; the test of comparison of two proportions in
> those is the one I already mentioned I used in my first post (except of
> course, that I did not subtract the second set of numbers from the first).
> It's a z-test similar to the comparison of two means. Is that the test
> you're referring to?
>
What I was calling a general test of association is often sloppily
referred to as a "chi-square" test, but there are many different tests
that are referenced to the chi-squared distribution. There should be
somehthing about testing for homogeneity in multiple proportions in your
texts. I tried to make it clear I was not referring to a test of trend
or implying any structure. Setting up an n x m table and calculating the
cell "chi-square" would be exceeding straghtforward on a spread***.
With n x m cells you would have (n-1) x (m-1) degrees of freedom. In
your data situation you have 9 degrees of freedom. (But who cares, the
test statistic is huge in your case.)

http://faculty.vassar.edu/lowry/vsfreq.html
http://davidmlane.com/hyperstat/chi_square.html

If you are missing an analysis engine and don't want to roll your own,
this page has links to web-driven calculators to this test:
http://members.aol.com/johnp71/javastat.html#CrossTabs

--
DW
.

Quantcast