Question

7. Assumptions underlying the repeated-measures t test 

Suppose a sleep researcher wonders whether playing classical music while getting ready for bed helps people fall asleep faster. She has 20 study participants sleep two nights in the sleep lab. For one of those nights, the researcher has each study participant listen to classical music for 15 minutes before getting into bed. For the other night (the control night), the room is quiet. (The researcher varies which night the study participants listen to the music so that 10 of the participants have the control night first and the other 10 have the classical music night first.) The researcher then measures the study participants' sleep latency-that is, the number of minutes it takes for them to fall asleep. 

At the end of the study, the researcher looks at the distribution of the sleep latencies and notices that some people fell asleep immediately, whereas others took over an hour to fall asleep. In addition, the differences in each participant's sleep latencies between the two nights are mostly less than 10 minutes, but for five of the participants, the difference is close to an hour. 

Would it be valid for the sleep researcher to use the repeated-measures t test to test whether playing classical music before going to bed shortens sleep latency? 

• No, because the sample does not appear to come from a normally distributed population. 

• Yes, because although there are deviations from normality, the total sample size is greater than 30. 

• Yes. 

• No, because the researcher's design uses independent groups.

Answer 1

since sample size 15 <30 , as well we can see that 5 participants have very high difference compared to other, we can not assume normality and therefore repeated measure t test can not be done.