Lecture 10: Response and Conditions Three Decisions

advertisement
Lecture 10: Response and Conditions
Three Decisions
Three Decisions
Chapter 4 looks at the Three
Decision discussed earlier
in more detail.
Response?
Conditions?
Experimental Material?
1
What is the response?
2
Diet Experiment
In the diet experiment the blood
pressure after 3 months was
used.
A better response might be the
change in blood pressure from
before the diet until after 3
months.
This seems like an easy
question but can be very
important to get it right.
3
Response Scales
4
Response Scales
It is not that important to
remember the names but to
realize that how we analyze
the response can depend on
the scale.
Nominal – named categories
Ordinal – ordered categories
Interval – distance between
values is meaningful
Ratio – there is a natural zero
5
6
1
Lecture 10: Response and Conditions
Example
Treatment 1
1
3
1
3
1
1
Example
When is it appropriate to
compute sample means and
run an analysis of variance
on data?
Treatment 2
3
3
2
4
3
3
7
Context 1
8
Context 1
Calculating a mean makes
no sense because the
numbers are really a
nominal scale, numbers
replacing names.
Response: Favorite color
1 = blue, 2 = brown, 3 = red,
4 = green
9
Context 1
10
Context 2
Response: Rating scale
A more appropriate summary
would be the proportion favoring
each color.
Treatment 1: 66.7% favor blue.
Treatment 2: 66.7% favor red.
1 = very unfavorable
2 = unfavorable
3 = favorable
4 = very favorable
11
12
2
Lecture 10: Response and Conditions
Context 2
Context 2
The response is an ordered
category scale.
4 is more favorable than 3.
3 is more favorable than 2.
Means and ANOVA are not
appropriate because the
distance between categories
is not the same.
13
Context 2
14
Context 3
Response: Temperature C
This is an interval scale
because 4 is 1C warmer
than 3 and 3 is 1C warmer
than 2.
Perhaps if it were a 5-point
scale with a neither
favorable nor unfavorable
being a 3, means might
make some sense.
15
Context 3
16
Context 4
Calculating means and
performing the analysis of
variance makes sense.
Response: Weight g
This is a ratio scale because
there is a natural zero.
4 g is twice as heavy as 2 g.
17
18
3
Lecture 10: Response and Conditions
Context 4
Comment
For our purposes, as long as
the response is numerical
(interval or ratio) calculating
means and performing an
analysis of variance makes
sense.
Calculating means and
performing an analysis of
variance makes sense.
19
Conditions?
20
Example
When deciding on the
conditions you want to
isolate the effects of
interest.
The text discusses an
experiment using mammary
ligation to reduce pain for
angina patients.
21
22
Another Example
Gastric Freezing
Gastric freezing for the
relief of pain from stomach
ulcers.
A patient with stomach ulcers is
sedated, a tube with a balloon on
the end is inserted down the
throat into the stomach, a very
cold liquid fills the balloon, the
liquid and balloon are removed.
23
24
4
Lecture 10: Response and Conditions
Gastric Freezing
Gastric Freezing
The very cold liquid is
thought to cauterize the
ulcers thus giving relief to
pain.
Most reports of the use of
gastric freezing did not
include a placebo, control
group.
25
26
Control Group
Results
A patient with ulcers is sedated,
a tube with a balloon on the end
is inserted down the throat into
the stomach, a body temperature
liquid fills the balloon, the liquid
and balloon are removed.
Patients exposed to gastric
freezing reported a reduction
in pain.
Patients exposed to the
placebo reported a reduction in
pain.
27
Results
28
Comment
The sedation, or patients
recognizing they were being
treated and so should feel
better, could be having an
effect on pain and were not
isolated from the gastric
freezing.
There was no statistically
significant difference in the
average reduction in pain
between the two treatments.
29
30
5
Download