Re: st: tukey/dunnett procedures

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

.

Thanks, Roger.

I would have thought there must be something else different between them, given Dunnett's test was meant for "one to many" comparisons of means.

I'll look again at both parts of these references to understand.

I asked because I was recently trying to understand if the ideas of Dunnett's test translate to other than a comparison of two means following a oneway ANOVA. I had a situation where I was using a full model for nonlinear regression and a set of reduced models, and one of the set was for a control condition and all the others were for treatment conditions, and the comparisons were all control to each treatment. I ended up adjusting the resulting pvalues using Sidak's or Holm's adjustment procedures.

I could not help shrug the idea that if Dunnett's has an advantage in the case of comparing one control to many treatments for means (well, typically less than 22 given the published tables), then is there not a procedure with similar advantages when comparing a set of models where one model is for the control condition and the others are for multiple treatment conditions?

> Equation (9) of Dunnett (1955) is a generalization to the t-distribution of the inequality demonstrated for the Normal distribution by Sidak (1967).
> Best wishes
> 
> Roger
> 
> 
> References
> 
> 
> Dunnett C. W. 1955. A Multiple Comparison Procedure for Comparing Several Treatments with a Control. Journal of the American Statistical Association 50(272), 1096-1121.
> 
> Sidak, Z. 1967. Rectangular confidence regions for the means of multivariate normal distributions. Journal of the American Statistical Association 62: 626-633.
> 

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