Statistics for unequal sample size?

4 posts / 0 new
Last post
sarakumar47
sarakumar47's picture
Statistics for unequal sample size?

Hi,

I have measured an antibody levels in two groups of animals after immunization.

Group 1 (n = 30) which was immunized and Group 2 (n =15) which was untreated used as control. To find the statistical significance between two groups, I think I have to do unpaired "t" test. Is it correct? If so, please let me know how to do it SPSS package 11.5? Otherwise, give me some suggestions on what kind of test I need to perform.

Thanks in advance,

Sami Tuomivaara
Sami Tuomivaara's picture
sarakumar47,

sarakumar47,

You're right about having to use unpaired test. If the control group was the same group of animals (pre-immunization), you'd have to use paired test.

But there are many flavors of t-test! In your case both sample size, and (most likely) variance of the antibody levels are different, and these need to be taken into account. Generalization of t-test, called Welch's test is the choice when sample size and the variance are different.

I can't help you with SPSS, but the equations look easy enough to enter in a spreadsheet and after getting the t-statistic and the degrees of freedom, you can proceed normally and look up the corresponding probability in a published t-distribution table, or calculate using spreadsheet function (in Excel it is called TDIST).

Cheers,

sarakumar47
sarakumar47's picture
Hi Suola, Thank you so much.

Hi Suola,

Thank you so much.

Biju
Biju's picture
Hi Sarakumar

Hi Sarakumar
You could do unpaired t test with microsoft excel also.
Biju Joseph