Paired T Test ???

#1
Hi

I am trying to work out which statistical test to use to determine whether there is a relationship between visual acuity (how small letters on a letter chart are) and contrast (how dark the letters are). I have used one sample of participants and tested both visual acuity and contrast on each participant. Will a Paired T-test do this?

Thanks
 

Karabiner

TS Contributor
#2
For paired t-test, you need measurements on the same scale.
What exactely do you mean by "relationship"? Maybe just a
correlation? Which correlation coefficient would be feasable
depends on how your variables were measured.

With kind regards

K.
 

Karabiner

TS Contributor
#4
Re: Paired T Test ?

If both variables were measured on an ->interval scale, then ->Pearson Correlation would be an option.
If one both variables were measured on an ->ordinal scale, or sample size is small (n < 20), then ->Spearman's rank correlation would be an option.

With kind regards

K.
 
#5
Re: Paired T Test ?

Thank you Karabiner.

I have carried out a Pearson Correlation (interval data, n=25) in Excel and it has come out with 0.179765513. Does this mean there is v little association between the 2?

I then did a t-test and it comes out with 4.90562E-57. Does this mean there is a high probabillity of no association.

Now I am even more confused?

Thanks again
 

Karabiner

TS Contributor
#6
Re: Paired T Test ?

unfortunately i don't understand what you presented as result of your t-test.
But anyway, you did not ask whether the means of the two variables differ, so
you do nor need a t-test. The degree of association between the two variables is represented by the correlation coefficient, which is indeed small in the sample
(if it actually was the coefficient r and not the p- value).

With Kind regards

K.