Up: Stat 3011
Stat 3011 Midterm 1 (Class Part)
All of this is summarized in the box on p. 25 in Wild and Seber.
On the first issue (what is the difference),
- In an experiment, the experimenter determines which
experimental units receive which treatments (ideally by randomized
assignment to treatment groups).
- In an observational study, which subject gets which treatment
is outside the experimenter's control. Perhaps the subjects or their doctors
choose. Perhaps it just happens (exposure to a toxic substance in the
environment, for example).
On the second issue (what is the implication)
- A properly designed and executed experiment can
reliably demonstrate causation.
- An observational study can only suggest possible causes.
It cannot reliably establish causation.
A properly designed and executed experiment is needed to follow up
results implied by an observational study.
The mean is
The standard deviation is
where
is the mean calculated in part (a), and
Thus
.
By the multiplication rule (using the assumed statistical independence)
So in order to answer this question we have to first answer the subsidiary
question: what is the latter probability? By the complement rule
Thus
The events in parts (a) and (b) of this problem are complementary, so
by the complement rule each is one minus the other
Curve A: skewed, long right tail, unimodal.
Curve B: symmetric, unimodal.
Curve C: skewed, long right tail, biimodal.
Curve D: symmetric, biimodal.
Up: Stat 3011
Charles Geyer
2000-10-25