In an experiment, you assign a treatment to one group and a placebo to the control through a random assignment with the researcher controlling the independent variable. Problems with observational studies Internal variable: could Y cause X? Is there some confounding variable? Ecological Fallacy: drawing incorrect conclusions about individuals based on aggregates Selecting on the DV: forgetting to let your variables vary Regression to the mean: are extreme values of Y due to chance Ecological inference: use of aggregate data to study the behavior of individuals Simpsons Paradox: ecological fallacy: do college admissions exhibit bias? When a relationship present in different groups is reversed when the groups are combined Separate: female acceptance higher but combined: male acceptance higher Observational design summary Cant always control x Solution: observational study Time series cross section But must control for alternative explanations Endogeneity probs Confounds Ecological fallacy Regression to the mean Fowler: iv: appearing on show; dv: amount of support the candidate gets Selection bias: non random factor affecting values of X. democrats not doing well so appear on show vs republicans doing well and then appearing on the show Zs: incumbency, party ID, donations prior to show Method: compare each politician who went to show and then match them with someone similar Findings for democrats: colbert bump but they are usually in trouble when they go Republicans: decrease in donations and no change in donations: mixed results Gender gap ended completely with the intervention and stayed with the control group. Midterm exam: be prepared to evaluate internalexternal validity of Miyake and Fowler Which is the bigger threat to the study The answer doesnt have to be right or wrong. It just has to have the right explanation.