What has been the reason for collecting the body fat data?
Of course you don't gather some data to see what you can do with it later.
That is not a statistically valid procedure to test hypotheses.
What is your expectation in this case, what is your hypothesis (H1)?
What is your design and why did you choose this design?
Why didn't you choose a separate control group?
How is the body fat measured? What unit?
Why is your sample (of paired data) only 10?
Did you perform sample size calculations in advance?
Why didn't you consult a statistician in advance?
You see, all kinds of questions that should be answered
before doing the experiment, before gathering data,
before being able to do any valid statistical analysis.
It is my impression that you do not have valid answers to all of these
questions. That means that you could do some statistical test, valid for the
type of data, but that it does not support or reject some hypothesis that
you did not have before gathering the data. It may only give an indication
of what to expect if you would repeat the experiment in a valid way.
I am sorry to say this but testing a hypothesis in this case seems to be an
illusion. It is also an illusion to think that statistical tools in the
hands of non-statisticians may perform miracles and magic.
Regards - Jim.
Jim Groeneveld, Netherlands
Statistician, SAS consultant
My computer shows magicant effects with miracustical tests on hypodata.
On Mon, 23 Oct 2006 11:35:15 -0700, illusion <mdasifg@YAHOO.COM> wrote:
>can someonehelp me out in.....what test should i use to test the
>hypothesis of effect of cycling,
>I have a dataset with body fat before and after cycling for a certain
>number of days for 10 subjects
>the data is normally distributed.....
> what sas procedure should i use to "test the cycling has no effect on
>body fat" -H0
>woould the test will also give me CI for the difference in mean before
>and after cycling???
>Thank you in advance