Paired t test- in obese and non obese females

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Paired t test- in obese and non obese females
Message posted by Dr Pooja (via 210.18.24.242) on November 21, 2001 at 7:52 AM (ET)

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Paired t test- in obese and non obese females
Message posted by Darius (via 200.23.217.10) on November 21, 2001 at 10:51 AM (ET)

Paired t-test means both samples are related (like before and after). Don't use it for non related subjects.

Re: Paired t test- in obese and non obese females
Message posted by Dr Pooja (via 210.214.101.228) on November 22, 2001 at 10:28 AM (ET)

so do you think a "t test" will be more appropriate?

Can there be any case when obese (n=20) and non obese (n=25) women are the sample that the paired t test is used?

can you suggest anyplace I can read lots of examples to get it clear?

Re: Paired t test- in obese and non obese females
Message posted by Darius (via 200.23.217.10) on November 22, 2001 at 4:05 PM (ET)

Yes, I think a "t test" will do the job.
>Can there be any case when obese (n=20) and non obese (n=25) women are the sample that the paired t test is used?
Nop, unless there exists a relationship between the two samples (y and z):
y(i) and z(i) related so the number of elements in both samples are equal (ie. somebody before and after a diet program).

Re: Paired t test- in obese and non obese females
Message posted by Dr Pooja (via 210.214.101.88) on November 23, 2001 at 2:21 AM (ET)

Yes,
the problem I have to solve is regarding the correlation of their cholesterol levels.

Do you mean if n is diff...we cant use the paired test
do i have to get the same number of obese and non obese women I mean?

Re: Paired t test- in obese and non obese females
Message posted by Mark Waters (via 128.243.220.45) on November 23, 2001 at 7:50 AM (ET)

You should not be doing a paired test. The people are not the same individuals. You would use a paired t test to analyse the effects of, say, a drug on reaction times. You would test reaction times for each individual both before and after drug application. In this case, you need to take into account differences between individuals in reaction times, so a paired t test is appropriate. This compares the effect of a drug on the mean reaction time before and after drug application, but the two samples are paired and must be treated as so.

HOWEVER in your case you have two independent samples: one set of obese women, one set of non-obese women. They are not related samples. To see if there is a difference you need to do a two-sample t-test which compares the means of whatever it is you are measuring of the two groups and sees if they are significantly different. You could do with more people if possible (bigger sample sizes).

Re: Paired t test- in obese and non obese females
Message posted by Darius (via 200.23.217.10) on November 23, 2001 at 10:30 AM (ET)

>the problem I have to solve is regarding the correlation of their cholesterol levels.

It looks as a correlation problem, not a test for means difference.

As Mark said they are not the same individuals (not the same "n", therefore, no before and after case), if you are willing to test means simple t-test will do the job, but correlation may be a better test for "cholesterol levels" against weight, age, etc.

Re: Paired t test- in obese and non obese females
Message posted by Dr Pooja (via 210.214.101.88) on November 23, 2001 at 2:21 AM (ET)

I really get the concept clear now. Thanks Darius and Mark, I didnt mean correlation as in the statistical meaning of the word.
Actually, the study was undertaken to assess the effect of a drug in lowering serum cholesterol levels in obese(n=20) and non obese women (n=25). So I thought probably we need a paired t test but now its quite clear that an unpaired t test is what I need.
Thanks for your time and effort.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.