T test for different groups over time


New Member
Hi all,

I know that paired t-tests are appropriate for the same group of people at several points in time while independent t-tests are for 2 different groups of persons. But what about when it is a combination of both (ie: looking at different people overtime?)

1) I want to compare 2 groups (smokers and non-smokers) on a continuous variable at 3 points in time (customers are interviewed every 12 months for 3 years). Those who are non-smoker at time 1 (1st interview) can “upgrade” to smoker at the 2ndor 3rd interview if they have taken the habit. However, nobody can “downgrade” from smoker to non-smoker.

Question: Do I need to use an independent or a paired t-test?

2) I remember reading somewhere that even if the DV in any t-test should be continuous, it will still be ok to do so with a dichotomous DV. Will anybody confirm?

Any help is very much appreciated.

Thank you!


New Member
my DV is a 5-point Likert scale on food diet : "on the following scale, how important or unimportant the following products have been for your diet"

Same questions are asked to all respondents in the 3 interviews


New Member
Yes! data collection has just finished!
Wave 1: 594 respondents in total (60% smokers)
Wave 2: 493 (55% smokers)
Wave 3: 309 (53% smokers)


New Member
Just to make things clearer: the same people have been interviewed over the 3 waves. ie: the 309 respondents in wave 3 have also been though waves 1 & 2 and no new comers were allowed in. Many people dropped during the waves though but that was expected.