Mark Siddall mes at
Thu Jun 23 18:21:43 EST 1994

In article <199406231522.KAA20445 at> afwagner at STUDENTS.WISC.EDU (Andrew Wagner) writes:
>As a beginner in statistics,  I have a question as to what method I should
>use to compare some data.  This is the project.  We are looking at serum
>marker levels in maternal serum (the triple screen).  Supposedly, these
>levels change with freezing and thawing.  We will be running 10 freeze-thaw
>cycles (aliquot, run the sample, freeze the rest, thaw, aliquot, run
>sample, freeze the rest...)
>Anyway, this will mean that we will have 10 data points for 10 patients. 
>We would be statistically comparing each numbered cycle (all of the first
>runs, all of the second runs, all of the third runs...)  They will be
>averaged for the 10 patients.
>What would be the best statistical method to use?  I am familiar with the
>student's "t" test.  Would ANOVA work better?  Which kind of "t" test would
>work the best?
What you want to do is run a Repeated Measures Analysis of Variance.  The 
standard ANOVA does not apply when your subjects (i.e., patients) are
the same across sampling periods (standard ANOVA requires independence
across samples).

Moreover, you will want to check your data for homoscedasticity (homgeneity
of variance) and for normalcy if you use a parametric Rep. Mes. ANOVA.

If they are not normal or homogeneous you'll have to transform the data.

If this all sounds funky to you, PLEASE consult a statistician at your
university.  If there is no stats dept, try some ecologists.


Mark E. Siddall                "I don't mind a parasite...
mes at                    I object to a cut-rate one" 
Virginia Inst. Marine Sci.                     - Rick
Gloucester Point, VA, 23062

More information about the Bioforum mailing list