POL 51 Lecture Notes - Lecture 6: Research Question, Confounding
Document Summary
How a confounding variable can influence how we interpret the bivariable relationship in terms of causality. Example: 1973 grad school admissions to ucb. Observations/cases: applicants to ucb graduate division in 1973 (n=12763) There is a correlation between gender and admittance. Maybe because of past educational inequalities, men tended to have better gre scores than women applicants. If you want to test for a possible confounding variable (z), try to look for a situation in which x changes but z doesn"t. To test: look at people with gre scores in the 90th percentile and above, and if the distribution of those people"s scores roughly matches the distribution of genders who were accepted/denied, then we can"t say ucb discriminated against women. Compare people with similar gre scores and see if. Confound: difficulty of admission to various departments. If it"s the case that women are systematically applying to departments with more difficult acceptance rates, we can"t say ucb discriminated.
