ANOVA with mean and SD?

[Michael Grossbach]

The balanced ANOVA requires equal numbers of parameters. If you'd place
your three treatment groups' data in one vector Y though, using a second
dummy vector of length(Y) to denote which group a data point belongs to
then ANOVA can deal with unequal numbers of data points (untested):
dummy = [ones(length(y(:,1)), 1); ...
  ones(length(y(:,2)), 1)*2; ...
  ones(length(y(:,3)), 1)*3];
Y = [y(:,1); y(:,2); y(:,3)];
[pval, f, df_b, df_w] = anova (Y, dummy);

If your data are not normally distributed use kruskal_wallis_test (for
one-factorial designs).

Michael

Wolfgang Schechinger wrote:
> Dear experts, 
> 
> I need to analyze 2500 parameters in 3 groups by ANOVA, means I want todetect which parameters vary between the groups. 
> 
> For every group I have a table with 
> 
> number of data points (may vary from group to group)
> mean
> standard deviation. 
> 
> (3x3 columns, 2500 rows)
> 
> As I understand the anova(y,g) function, one needs to supply the raw data and there should be the same number of data points per colum at least for each parameter.
> 
> Is there a way to adapt the function to my data set, or is there another function one could use?
> 
> What can I do when my data does not follow a normal distribution?
> 
> Many thanks for your help!
> 
> Wolfgang
>