Different degrees of freedom when reporting multiple T-tests

Please refer to Table attached. This study compares those without a phobic disorder (n=83) with those with a phobic disorder (n=16) on Temperament scores.

What I am confused about, is the reporting of the degrees of freedom: shouldn't the degrees of freedom all be the same number (i.e., n - 1)?

And if so, does this mean the authors have used a different sample number for each independent t-test? What do you think is going on here in this Table? Any help appreciated!


Less is more. Stay pure. Stay poor.
Well, is it possible there is missing data or not everyone was positive for the category?

Also I thought for unpaired t-tests the df = n1 + n2 - 2, so they have non-integers as well for df. Off hand I am not recalling why that would be - perhaps Welch's unequal variance test?
Thanks for your response hlsmith :)

I'm not sure why the df would have a decimal place either. They haven't reported anything in their analysis section about Welch's unequal variance test. They haven't really reported anything.

Could you please explain what you mean about "missing data or not everyone was positive for the category"?

Thank you!


Ambassador to the humans
Yeah the unequal variance version of the t-test has its degrees of freedom modified to deal with the unexactness of the test. So what you're seeing isn't too unexpected.