3: Evaluating Frequency Claims Flashcards
Interrogating construct validity
Empirical reliability:
- test-retest: stable variables
- inter-rater: observational
- internal: self report
Subjective:
- have face validity?
- tap into all aspects of variable (content)?
Empirical evidence construct:
- quantitative or categorical(criterion validity)?
- correlate with measured (convergent) or with measures of other variables (discriminant)?
Interrogating external
-attend to method of recruitment and sampling
-potential for sampling bias? How were recruited?
•was convenience?
•other techniques?
•do recruiting cover targeted population?
Potential for volunteer bias? Who is particpatents?
•self election result in non-response bias?
Problems with telephone sampling
-using phone book =sampling bias
•individuals with and without landline diff
-even including cell phones =low response rates and introduce (non)-response bias
•those who answer diff from those who don’t
Internet research
-nonrandom sampling:
•sampling bias: internet access, know how to use, internet-savvy
•volunteer bias: self selected opt-in
-from internet and not internet may be diff
•may be a problem for sensitive or hot button issues
Interrogating stat validity of frequency
- proportion at specific value of variable (rate) and margin of error
- mean value of specific variable (level) and standard error
•large sample size: smaller margin of error
-reflects more confidence in population
- sample size formulas
•use an economic sample size
-even for national popualtions, should not large
-only enough to ensure acceptable error
Sample size doesn’t equal external validity
- larger samples not neccassarily more representative
- reflects degree to which sample was randomly selected and looks like POI not sample size
-once external established, increasing sample size decreases margin of error
•don’t nothing for biases (sampling or response)
1936 presidential election
Landon v Roosevelt
-mailed block ballots to a mailing list of 10 mil American voters
•1/4 voting population
-2.3 mil responses, Landon (R) would win
-FDR won handily
Dewey v Truman
-Gallup poll trained interviewers to meet demo quotas
•pollster chose who to interview
-dewy defeats Truman