Suitable statistic test

#1
Hello,

I have been reading a scientific journal and I want to check the statistical significance between two groups in their results. I have the mean, the n number and the SD for each group but not the actual data. An un-paired student T-test would do the job but it needs the whole range of data which as i said i don't have. Any suggestion for a statistical test which I can do, using the values I have?

thanks
 
#2
Hi,

What are your assumptions?

Are both n values > 30?

What is the distribution if they are small samples?
You could do an F test sigma1 = sigma2.
Ho: s1/s2 = 1 vs Ha: s1/s2 =/= 1, put largest s as numerator. Numerator and denominator df are the respective values of n-1.

You could do a T test for small samples mean1 = mean2.
first find out if upper, lower or two tailed test
assuming sigma1 = sigma2 form pooled variance
Ho: mean1 - mean2 = 0 vs Ha: mean1 - mean2 >0; or <0; or =/= 0. df = n + n -2.

Or Z test for large sample.

Peter