# Testing statistically significant difference between samples

I have 1 variable measured for 8 samples. so, for e.g. height of 8 men. I want to say something about: if height of men is significantly different amongst them. How can I do this?

Without any reference theoretical mean, I can not do a simple t-test.

-
As is, this question is unanswerable, as far as I can tell. You have the height of 8 men. If that's all you have, you cannot do any statistical testing. You could find the mean and standard deviation and so on. "Statistically different amongst them" is not a hypothesis, so it can't be tested. Also, with only 8 people, even if you had another variable (age, race, political party, whatever) it would take an absolutely huge effect size to find statistical significance. – Peter Flom Dec 16 '12 at 14:54
I changed "parameter" to "variable"; probably "sample" should be "subjects" as well. – Peter Flom Dec 16 '12 at 14:55
Thanks Peter. Indeed, I can add 2 more variables to the height- lets say weight and age. But, is there anyway, i can say that the samples are different or not? Can i do something with mean or median? any suggestions will be helpful. – abhishek Dec 16 '12 at 14:56
You need to formulate a hypothesis. "Samples are different" is not a hypothesis. If you said "these heights are different from the heights of XXXX" you could test it. XXX could be taken from some general population for which you know the average height. Or you could just take a number (e.g. "these heights are not equal to 175 cm". – Peter Flom Dec 16 '12 at 15:02
thanks again. I got the point. – abhishek Dec 16 '12 at 15:20