On Mon, 20 Jan 2003 23:13:11 -0800, "Jeremy Bauer" <[EMAIL PROTECTED]> wrote:
> Good Day, > > I have a problem that seems like it should be easy, but always seems to > bring confusion to others as well and was hoping someone might be able to > help. Usually when I first describe the problem people are "Oh, you > just....." and they complete an answer then say, "But, wait...hmmmmm" > afterwards giving me no confidence. Here's the problem: > > I have 10 subjects, each performing 2 different activities 10 times. I am > looking at the difference in maximum left foot-ground forces during running > and during landing from a jump. So, I measure foot-ground forces for I assume that you refer to the 'maximum force' during one jump, and you are interested in comparing the average of those scores. > subject A 10 times while running and 10 times while landing from a jump. I > want to know if the maximum measured forces are different between tasks. > > I've performed several studies in the past showing that foot-ground forces > do not change as a function of trial. However, I need to collect 10 trials > due to the variance that exists in foot-ground force measures. In the past, > I have calculated the mean foot-ground force for each subject in each task, > then just used a paired t-test. This seems reasonable, however while no one > has ever told me this approach is incorrect, no one has told me it is > correct. I basically just don't have much confidence in that approach. > Especially if the subjects vary a lot in size, it might be more meaningful to consider the log of the forces, or the ratio of the (average) forces. Other than that, the approach that you describe seems perfectly reasonable. A sensitive listener, though, will probe to ask WHY you don't have much confidence.... Is there something more to consider? Are there "odd" data points, somewhere in the tangle? How systematic are the results? -- Do you want a report in a different fashion? -- I think I might like to hear that, "A ranges from 75% to 110% of B. On the average A is 90% of B, and the interquartile range is 85% to 95%." > I would greatly appreciate either some confidence boosting or someone that > can point me in a direction that might be a better approach for this > problem. If you are sure that trial-number is irrelevant, you keep the problem simple by averaging at the start. Don't explain the data in that much detail unless it matters. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
