"Gap between rich and poor has grown in more than three-quarters of OECD countries over the past two decades"
|
Hand in next time: + + + + + + + + + + |
Read, to discuss
p. 233, 9.43 quick randomizing p. 234, 9.47 explaining medical research p.233 9.41 prayer & meditation (clarification:
they help the person praying; careful experiments to see if they help a
person prayed for have not shown positive results.) p. 232, 9.38 spine fractures block design. You lack the information to make a complete design (i.e. how many women at each hospital.) Sketch in what you can. |
Optional Think
about this one... .Postpone . p. 226, 9.14 matched and not, more practice |
"Countries with a wide distribution of income tend to have more widespread income poverty. Also, social mobility is lower in countries with high inequality, such as Italy, the United Kingdom and the United States, and higher in the Nordic countries where income is distributed more evenly.
In most OECD countries, around half of poor people are
better off and move above the poverty line within three years. This
figure is the highest in Denmark and the Netherlands. Income mobility
means that people who are persistently poor make up less than 2% of the
population in these two countries. But
persistent poverty is much more widespread – 7% of the population – in
Australia, Greece, Ireland, Portugal and the United States.
(Finfacts) http://www.finfacts.ie/irishfinancenews/article_1015037.shtml
E.g. 2 headache medications, in combination?
A two-factor experiment, each with 3 values (levels).
9 possible treatments.
Factor A: Aspirin: values: None, 500 mg, 1000 mg
Factor B: Caffeine: values: None, 50 mg, 100 mg
Response variable: reported pain relief
Lurking
variables: Control--how?
Nothing except experimental treatment should differentially affect
response.
Compare responses
under several treatments, look at differences.
Placebo effect: a positive response to a "sham" medical
treatment
"Control group" Group that gets the "baseline"--"null"--
"none" or "placebo" value of the factor.
Always: trying to make everything "the same"
except for our treatments, to try to eliminate
confounding/lurking variable effects.
.New today.
How to get groups "just like" one another? Randomize who goes
into which group. (Usually our batch of experimental units
is not a random sample
from the population of all individuals--volunteers, etc.)
Randomized comparative experiment : Diagrams
of design, Moore pp. 218-19: shows where
randomizing happens, how many to each treatment, what
the treatments are.
Completely randomized: all exp. units allocated at random among the
treatments. Applet or random # table.
Use enough subjects for each treatment so that you can "average out" (and measure) chance variation in the subjects.
Principles of designing an experiment (p. 221) See above:
More problems, cautions:
Placebo and biasing effects can result from
expectations. "Blind", "Double blind"
Did you forget to measure something that might be a lurking
variable?? Measure everything you can...
Lack of realism: Do sociology, psychology experiments
generalize to "real life?"
--Subjects are not a random sample from the population. (Most
psychology "facts" were based on studies of Ivy League males, before
1970's.)
--Ethical questions...Milgram. Whole section BPS4e, pp.
235-242
New material:
Statistical
Significance p.221: An observed effect so large that it
would rarely occur by chance (assuming no real difference in
treatments) is called "statistically significant".
"So large", "rarely", "by chance" will be defined and quantified in
Ch. 6.
Example: Suppose 95% of the
subjects had their headaches cured by treatment 9 and only
25% by treatment 1 (placebo). IF the medicine in fact did
"no good" that would be a very unlikely outcome (at
least if we have "enough" people in each treatment). So we
will say the difference in headache cures between treatment 1 and
treatment 9 is "statistically significant" and be inclined to believe
the medicine "works".
Example: Gallup says Obama "significantly" ahead: difference they got
is so large that they would rarely see such a difference in a sample if
indeed the two candidates had the same support. (What's "rarely"? They do samples all the time.
Rarely = 1 in 20? 1 in 100? 1 in 20 is common usage, 1 in
100 also...)
= = = = =.Start here Friday.
= = = = = = =
Fancier Experimental designs
(not "completely randomized") Control extraneous
variability by pre-sorting individuals into homogeneous
"blocks". Do usual design on each block. (BPS4e pp. 224-226)
Matched pairs: To compare
Control and experimental treatments (i.e. 2 values
only)
Sort experimental units into "matching"
pairs. One member of pair gets control, other gets
experimental.
Randomize which.
Compare within pair, then
summarize all comparisons.
Common: Do the control and experiment to same
individual (matched with self = "self-paired"). (Randomize order)
Are right feet bigger than
left feet? (not an experiment) Sunburn
salve experiment?
Aside: Sampling data, "longitudinal
study" following same people through time.
Works like matched pair to control variability.
Block design: Sort experimental
units into "Blocks" = groups homogeneous on potentially
confounding variables
e.g. M/F, age, income, weight, fruitflies wild
or curly-winged. (No randomization here yet.)
Within each block, randomize the treatments. Compare
results within each block, then summarize all results.
(Matched pairs is a special case of block
design--each pair is a tiny "block".) Diagram p. 226
| Sievers home | Math151-Fall08/Dayf23.htm | 3pm | 10/22/08 |