paired t-test assumptions

[kalbright36]

I know this test is used for two groups and involves a normally distributed, continuous variable--and it involves the difference in means. Does the assumption include equal variances as in the case of the independent t-test? Would that be the case for one-sample t-test, too?
If it sounds like I don't know what I'm talking about--you're right!
Thanks very, very much!

 

[JohnM]

The paired t-test actually tests hypothesis that the average difference or delta between paired or matched individuals is different from 0.

Ho: d = 0
Ha: d <> 0

So, to answer your question, no it does not assume equal variances - it doesn't even come into play. Same for the one-sample t-test as well.

 

[kalbright36]

t-tests

Is an unpaired t-test the same as an independent t-test?
Thanks some more!

 

[JohnM]

Yes. Independent groups means that the items or individuals in the two groups are sampled in a manner that is totally independent of each other.

 

[kalbright36]

t-test again!

I'm looking at an article. There is a table showing the mean with std error of the mean for each variable (pre- and post-intervention). Can I check to see if they are normally distributed using these values (i. e. divide each mean by SE and look up in z or t table)?

 

[JohnM]

Dividing the mean by the std error of the mean won't tell you anything...

You'll need the individual data points to determine if a data set is normally distributed or not.