Re: st: repeated measures

["Airey, David C" <david.airey@vanderbilt.edu>]

.

Explaining a complicated experiment on a listserv is not possible, really.

Because the outcome appears to be "contaminated" and "not contaminated", I'd
suggest using ANOVA is the wrong model. ANOVA wants a continuous outcome. You 
have a binary outcome.

A "logistic regression" model is good for binary dependent variables.

If the data are repeated measures, you could look into using
Stata's xtgee command with options for logistic regression. Look at
a couple examples in the Stata PDF help for xtgee.

One thing that may be of concern is whether your hypothesis is about 
equivalence or superiority. You might Google for that distinction.


> Thank you Dave --
> 
> sorry to be so obtuse. the design is a prospective cohort of healthy volunteers.
> 
> the test is a unrinalysis, on which is re reported five relevant
> indices: nitrates, leukesterase, WBCs, epithelial cells and bacteria
> 
> we collect these twice, by different methods, 1 and 2, from each subject
> 
> The outcome variable is whether the sample grew a out bacterial
> evidence that the collection was "contaminated"
> 
> So, in wide form, I have:
> 
> subjid nit1 leuk1 W1 epi1 bac1 contam1 nit2 leuk2 W2 bac2 contam2
> 
> in long form I have a "collect" variable, 1 or 2, and then nit, leuk,
> W, epi, bac and contam
> 
> I want to analyse all 80 specimens for the odds that a given index
> result will predict contamination, accounting for repeated measures on
> both the predictors and the outcome of "contaminated"
> 
> I hope this clears it up!
> 
> Thanks for you help,
> 
> Harrison


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/