MIKEY: no problem - first let's not do pairwise t tests, OK? we would have RA Fisher turning over in his grave, and that would not be good. so, it appears you have three independent groups and you want to test the null hypothesis that the three population means are equal. you need the sum of squares between and within groups. then you can get the mean squares by dividing by DF, form the F ratio.
SS between is summation of squared deviation of each group mean from the grand mean each weighted by the sample size. DF=k-1 where k is number of groups (3). SS within is (n-1)*variance for each group then summed over the groups. DF=total N minus k. this mean square between/mean square within follows the F with k-1 and N-k degrees of freedom. this should do it. Michael Steiper <[EMAIL PROTECTED]> wrote in message news:<b276rl$jd8$[EMAIL PROTECTED]>... > Hi, > > I've been messing with a group of data that I don't have raw valus for. I > ONLY have means, standard deviations, and N. Generally these three values > are for three classes. I want to do either t tests (pairwise [i know it's > not the best]) and/or ANOVA. The Ns are not paired. What is the proper > formula for a t test? is there any way to do an ANOVA? If no ANOVA do i > need to correct my alpha vals for the pairwise t-test? > > Thanks, > > Mikey . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
