Pingu wrote: > > > This isn't the best way to evaluate the configurations (though it is slightly > > better than doing so on the basis of no data at all). It depends too heavily > > on one person's characteristics (e.g., prejudices, hand/finger size, previous > > experience and so forth). > > Ok - but is it not acceptable to run a test which measures which > condition is significantly fastest FOR ME?
Sure you can run the test. What would you do it is non-significant? If you would use those data to determine the next stage of your study in any case, I'm not sure the test helps. (Assuming you are already committed to picking a subset of these configurations). In either case (significance or non-significance) the average (e.g., mean or median) and the dispersion (s.d., s.e., range) will be determining you choices. If I've misunderstood and you'd have time to go back and collect more data then a significance test _might_ be of use. Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
