| |
 |
|
|
Science Forum Index » Statistics - Math Forum » T-tests
Page 1 of 1
|
| Author |
Message |
| Jcole |
 Posted: Wed Mar 14, 2007 10:01 am |
|
|
|
Guest
|
|
| Back to top |
|
| Vassili Pastushenko |
| Posted: Wed Mar 14, 2007 10:23 am |
|
|
|
Guest
|
|
| Back to top |
|
| Ray Koopman |
| Posted: Wed Mar 14, 2007 4:12 pm |
|
|
|
Guest
|
On Mar 14, 1:01 pm, Jcole <jscole...@toadmail.com> wrote:
Quote: Hello:
I hope someone here can help me. I've posted to other stat discussion sites, but nobody seems interested in replying.
We have a data set consisting of 6 groups, with n-numbers= 2,3,2,1,3,1, respectively. We are attempting to run ANOVA, and Dunnett's/Bonferroni. Our stat software bombs out because of the n=1 sample sizes of two of the groups. Our analyst is adding "dummy" values to these groups that are equivalent to the single sample in each one (i.e. if the value is 23, the input data is 23 and 23 for that group). That way, the mean is retained, but now the n=2, and the program runs. Somehow, I don't think that's a legitimate fix. Is there software available that will handle n=1 sample sizes for this, and/or are there more appropriate statistical tests to use when this occurs? Thanks.
Look around for software (StatXact?) that will do a randomization
test, that compares your obtained between-group variability (or
within-group variability -- it doesn't really matter which) to
its sampling distribution under random partitioning of your 12
observed values into 6 groups with the same sizes your groups have. |
|
|
| Back to top |
|
| Jcole |
| Posted: Fri Mar 16, 2007 7:21 am |
|
|
|
Guest
|
| Thanks for the replies. How about a one-sample t-test? I should have specified initially that the first group is the control, and we're looking for differences between treated and control. |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sun Oct 12, 2008 10:53 pm
|
|