KLMED8008 Repeated measurements
The exercises are voluntary and do not need to be handed in. However,
they are revelant for the final one-week take-home exam. Lecturer are
present to
provide assistance and guidance.
Exercise 1
1. Textbook, Exercise 1.2 (anorexia.dta, p 41 in 2nd ed., p 61 in 3rd ed., p 64 in 4th ed.)
2. To estimate sample sizes in an experiment in which two groups are
compared and the outcome is adjusted for baseline, run the following
Stata commands:
sampsi 55 60, sd(15) pre(1) r01(0.55)
sampsi 55 60, sd(15) pre(1) r01(0.35)
The only difference is that the correlation between baseline and
outcome variables differs (0.55 vs. 0.35). You may
wish to try other values as well.
Verify by inspection that the best (more effective) analysis strategy in each case is "ANCOVA".
Try to explain why the "change" strategy (i.e. paired t-test) is
ineffective - and in fact worse than the "post" strategy - when
correlation is low.
3. Textbook, Exercise 2.7 (birthwt.dta, p 89-90 in 2nd ed., p 119-120 in 3rd ed., p 130 in 4th ed)