How to do paired t test with mean and sd?

I would like to do paired t-test with mean and standard deviation and the number of sampling.

for A:mean =8.37 sd=0.53
B: mean=8.18 sd=0.16 number of samples for both A=B=44

Thanks.

Answer 1

[Star Strider]

To do it with only the means and standard deviations, you would have to write your own functions to calculate the one- and two-tailed t-probabilities, but fortunately with MATLAB that is not difficult:

Amean =8.37; 
Asd=0.53;
Bmean=8.18; 
Bsd=0.16;
N = 44;
v = 2*N-2;

tval = (Amean-Bmean) / sqrt((Asd^2+Bsd^2)/N);       % Calculate T-Statistic

tdist2T = @(t,v) (1-betainc(v/(v+t^2),v/2,0.5));    % 2-tailed t-distribution
tdist1T = @(t,v) 1-(1-tdist2T(t,v))/2;              % 1-tailed t-distribution

tprob = 1-[tdist2T(tval,v)  tdist1T(tval,v)]

tprob =
       0.0253      0.01265

So your groups are significantly different at alpha = 0.05.

The functions are correct, but you may want to check my ‘v’ and ‘tval’ calculations to be certain it they are correct to your experimental design. If they are not, change them appropriately.

[Rita]

You are amazing! Thanks

[Star Strider]

Thank you! My pleasure!

[Aviram Shemen]

Hi,

Is there a way to do the paired t-test if I don't have the samples sizes?

(I'm trying to compare two results of a Cox regression model using coxphfit. So I have two means - exp(b1) and exp(b2) and two SE - se1 and se2)

Thanks in advance

[Jeff Miller]

This old answer is not correct for a paired t-test. In that case the tval computation depends on the standard deviation of the difference scores A-B. You can compute the required value from Asd and Bsd if you also have the correlation of A with B.

[Aviram Shemen]

Thank you!

Answer 2

[Thorsten]

If you have the Statistics Toolbox, you can use

ttest(A-B)

[Rita]

Thanks a lot.I have this toolbox but can ttest function work with mean and sd without samples? Thanks again

[Thorsten]

No, you need the samples.