t-test for sample comparison
kr1 at PGSTUMAIL.PG.CC.MD.US
Mon Nov 17 09:43:11 EST 1997
Perhaps you could utilize a nonparametric statistical procedure
such as the Mann-Whitney U. If your null hypothesis states that there is
no significant difference between control and product, the distribution
of U scores can be used for testing hypotheses in the same manner as t or
x2 distributions, without the necessity for meeting the assumptions placed
upon T. I hope this helps.
Karl J. Roberts
On 15 Nov 1997, Rick Parcels wrote:
> I have to perform the preparatory test to the microbial limits test (USP)
> on all of my company's final products. The method I would like to use to
> validate the test is as follows:
> Broth/Product mixture is inoculated with 10-100cfu/mL. Mixture is plated
> 10 times using 1.0 mL. Counts are taken.
> Broth/Water mixture is inoculated with 10-100cfu/mL. Mixture is plated 10
> times using 1.0 mL. Counts are taken.
> These results from these two mixtures (Product and Control) are compared
> using a t-test.
> Does anyone know if the t-test is the appropriate statistical method to be
> using to comapre the results? I learned in school that the t-test can only
> be used if the larger variance divided by the smaller variance is < 2. My
> ratio is much larger than two. I can't find any other appropriate method
> of comparing two population means (or would this be a sample mean?).
More information about the Microbio