Running ANOVA with only means, variances and n for data?

#1
Hi All,

I have problem where I want to run an ANOVA on four groups with unequal variances and unequal sample sizes but cannot ANOVA in the normal fashion.

I have to calculate the mean and variance for each group by hand and input that directly into SPSS rather than run the analysis on all the cases for the variable in question.

Is their a way I can manipulate SPSS through syntax to do this. I am particularly interested in running the Brown-Forsythe test. I have actually made all the calculations by hand but do not trust the sources I used to calculate the MSE and adjusted DF. So I want to check it with SPSS.

The variable I am using is actually the length of a vector so I cannot simply take the mean of the three components making the vector nor the standard deviation of the length of the vector. Hence I calculate the mean and variance by hand. It is to do with Somatotypes and anthropometry for anyone interested.

Any advice would be appreciated.

nick
 

mp83

TS Contributor
#2
I think that unequal variance between groups is a problem for ANOVA. You can either use a robust statistic for the test of equal variances or move to MANOVA.