On 14 Oct 2003 05:13:37 -0700, [EMAIL PROTECTED] (Baldy, Richard) wrote: > I am returning to the September thread that dealt with Ancova and its > homogeneity of slopes assumption. > > In Excel I wrote a "power" program to determine the likelihood an > experiment with one experiment factor set up in a completely randomized > design with one covariate and with a set number of replications would > find preplanned treatment contrasts significant. The program also > determines the likelihood that the assumption of homogeneity of slopes > would be violated. I find that as the number of replications increases, > so does the likelihood that the assumption would be violated - with 20 > reps about 90% of the simulated data sets yield a p value for > homogeneity of slopes less than 0.05.
If I follow this correctly: You have done a Monte Carlo randomization. < something unintelligible, about a test and replications. > In the MC, 90% of the simulated sets yield a p-value of less than 0.05 for something that ought to be random. Well, that is just wrong. I am puzzled, that the POSTER does not recognize a problem shown by that qualifier, "as the number of replications increases, ..." . Does that say, "When I increase the size of the problem, I get results that are increasing different from what they ought to be..."? - That seems to me to be a sure-fire indicator of a programming bug. [ snip, most ] > selected by the experiment planner. For each treatment (diet) the > covariates and responses are arranged so the highest covariate value is > matched with the highest response. The reason for this arrangement is > that one would expect, say, the largest animal to have the highest daily > gain. well, yeah, sometimes; to some extent, big goes with big. I don't know, for sure, what that has to do with the earlier problem with the simulation, but it surely does seem like a unique way to simulate a correlation. If you are simulating a correlation in excess of 0.99 (by having perfect rank-order correlation), then , YES, I would not be surprised to get 'non-homogeneous' as a test result. Okay, fix that by learning how to put in a specified correlation, into a simulation. Is there still something odd in the results? [ ... ] -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html "Taxes are the price we pay for civilization." . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
