t-tests or mann-whitney??

[zxcvbnm]

Hello,
I have a n=12 sample (and a matched-sample of 12)... I was planning on using t-tests, but now reconsidering it. Would it be correct to use a mann-whitney test to compare the means of the 2 samples (because of the sample size)?

I did K-S test to check for normality and no-significance was noted- so I'm assuming my data distributions are normal... was this the correct way to check for normal distribution. I tried looking at histograms to eye-ball it, but I don't get it!!

all things equal, are there disadvantages to using Mann-Whitney U rather than t-tests?

I'm trying to figure out the most appropriate way to analyze my data... any advice would be appreciated. Thanks.

[jamesmartinn]

From the sounds of it, you at least have the normality assumption satisfied. You should eye-ball the histograms or use Q-Q plots for each group too, however.

What about homogeneity of variances? A Levene's / Bartlett test will let you know if you have heterogeneous variances or not.

Mann-Whitney is a non-parametric alternative to the independent samples t-test. So unless your data is in the form of ranks (i.e, measured on an ordinal scale) or you have severe violations of the assumptions of normality/equality of variances, you should use the T-Test approach.

[wipeout]

If you doubt about normality, you can use permutation test. This kind of test doesn't assume normality or independence, just variance homogeniety.
You can find it on package "coin" in R/S-Plus