Professional Documents
Culture Documents
seems more plausible? Note: this is a preview of thinking about sampling distributions
Surveys
Subjects fill out questionnaires Choices:
Random sample
Same characteristics as the population list the whole population draw random numbers to select a subset (sampling without
replacement)
being selected
correspond to target population nonresponse bias respondents with a certain characteristic are more likely to not fill out a survey
(e.g. US Census undersamples blacks vs. other
ethnicities)
available to a wide audience online, those who fill it out may have a bias
e.g NRA survey ABCs online survey of addiction to the internet
Kinsey reports
studies of human sexuality widely used to support the claim that 10% of the
male population is homosexual (also an example of a persistent statistic, that was never explicitly made by Kinsey) criticism:
over-representation of some groups in the sample:
25% were, or had been, prison inmates 5% were male prostitutes.
non-response bias
impossible to you that the Nazi extermination of the Jews never happened?
22% - possible 12% - didnt know
that the Nazi extermination of the Jews never happened, or do you feel certain that it happened?
1% - possible 8% - didnt know
nonresponse: people who could not be reached or declined to take the survey
sampled population
whose birthday falls next What kind of sampling bias occurs with this strategy?
people in large households underrepresented young people underrepresented
exponential growth
assume the # of children gunned down in 1950
# of children gunned down over time (assuming doubling every year since 1950)
500 # kids gunned down 0 1950 100 200 300 400
1952
1954 year
1956
1958
# of children gunned down over first 20 years (assuming doubling every year since 1950)
0 e+00 1 e+05 2 e+05 3 e+05 4 e+05 5 e+05 1950
1955
1960 year
1965
1 e+00 1950
1 e+04
1 e+08
1 e+12
1960
1970 year
1980
1990
2000
Yearbook--1994 does state: "The number of American children killed each year by guns has doubled since 1950."[1] How does the difference in wording change things? Are there phenomena that truly are exponential?
growth of wikipedia
population projection
tuned exactly the way they are (out of all the possible parameters) How did everything balance out exactly right to allow for life to exist? (weak) anthropic principle, a truism:
conditions that are observed in the universe must
Experimentation
Control one variable measure effect of another Randomization
taste tester tastes soda in random order survey questions appear in random order
Blindness
subject does not know which treatment they are receiving
(placebo vs. new medication, or diet vs. regular) double-blind: experimenter making observation does not know which treatment the subject received
e.g. ESP research
UN Survey
Please complete the survey individually without
observational studies
claims made by test-prep organizations students taking a test-prep class increase their SAT
scores by 100 on average is this proof that test-prep classes are the cause of the improvement? What are possible lurking variables?
After
freakonomics
which studies were observational? which studies were experiments? how did the studies minimize sampling bias?
freakonomics: studies
bagels crime stats parenting school choice cheating teachers sumo wrestlers real estate agents
summary
3 types of studies surveys observational experiments in all of them, we have to consider population sampling frame non-response bias resulting from survey/experimental design
and sampling