Stats doesn't seem to be real popular, and I am dead rusty, but...

I understand what you want to say, but did the data bear it out?

Try to find relevant statistical testing procedures under 'non-parametric analysis of variance'. Mann-Whitney's test is fine for comparing 2 sets of measurements. If you have more, you'd better compare them all.

You have to compare all the different treatments using ANOVA to be certain that the populations are seperable. If they are seperable populations then you can apply your other tests to the populations you want to. You're on the right track.

Finally, I want to say that "balance is related with disease, but not aging."

Are you seeking information or trying to "spin" a pre-conceived idea? I have gotten to the point that I can tell you that one's sense of "balance" deteriorates with aging even for the otherwise healthy. When I play basketball now, a "spinning jump shot" is a real adventure. (Let alone the twinges in the knees after the game.) And I don't even want to think about riding that 3-dimensional tilt table from my USAF days.

You don't mention the type of data you have collected, but usually the hypothesis and the measurement of dependent and independent variables are identified in the design of the experiment before the data is collected for ANOVA.

I suggest you study up on what model and type of ANOVA you can employ with the data you have taken. Depending on what results you are looking for, you may consider retesting to gain more or more significant data (been there!). Don't forget to report the confidence interval for your conclusion.

Good Luck!

What are we comparing? Differences in Averages, or differences in standard deviations?

The test you used (I called it the wilcoxon signed rank test when i had stats) compares the median to an hypothesized value. Where did you come up with the hypothesized value for 'balance' for these groups? How large are groups? hospitalized people are biased data set.

This test is valid only if the population the sample is drawn from is "normal."If you have two genders in each group, you will have bi modal, non normal data.

Personally, if I were interested in your question, i would probably be more interested in the dispersion of the balance data than its central tendency.

Then i would investigfate what was difference between high and low deviation in each group- since all would be same age, the high std deviation people would be rich in clues as to what disease or other factors are to blame.

Walk first. Then run.

milo

let me know where i can use this stastical methods?