Natalie:
There is a technique called "boostraping" for resampling from
your data set to get a more stable estimate of values. You may
want to check your local stat package to see if that is
available.
For class assignments I use "lo-tech" techniques like Jim and
create pseudosubject scores by jittering the data with calls to
a random number generator such that the mean jitter effect is 0
and the range is some reasonable value, depending on the
question.
Treatment effects may be simulated by adding a jitterized
constant.
Ken
On Tue, 07 Nov 2000 15:33:53 -0500 Nathalie Cote
<[EMAIL PROTECTED]> wrote:
> Suppose you were in a Methods or Statistics class in which you had to design
> a study, create the materials, collect some of the data, and then use the
> data from this partial sample to simulate or fill in fictitious data for the
> rest of the sample.
>
> Let's say, for example, that you've collected scores on self-esteem and body
> image instruments from 10 college men and 10 college women, and you want to
> use those data as the basis for generating fake data for 40 more women and
> 40 more men. The data will then be analyzed for correlations between the
> measures and differences between men and women on each measure and on the
> correlations.
>
> One thread, of course, might be to discuss why anyone would want to do this.
> Let's assume for the moment that it's done for good reasons. What I am
> interested in is, how would you go about simulating the data?
----------------------
Kenneth M. Steele [EMAIL PROTECTED]
Dept. of Psychology
Appalachian State University
Boone, NC 28608
USA
