PUBHLTH 1 Lecture Notes - Lecture 11: Bran, Pancreatic Cancer, Recall Bias
Document Summary
Problems and limits of epidemiology problems with studying humans: intervention study: Subjects may not follow prescribed behavior throughout study period. Sometimes hard to isolate which of many factors are responsible for health differences: cohort study, case-control study: Control group may not be truly comparable also errors in reporting or recall: for all studies, must worry about differential dropouts. Sources of error: news reports (oat bran prevent cancer, to make no difference, random variation (results by chance, confounding variables (coffee, smoking, pancreatic cancer, bias selection bias ( heart disease patient, control digestive problems, dietary fat) Factors that lend validity to results: strong association, dose-response relationship, known biological explanation, large study population, consistent results from several studies. Students of journalism are taught that a good news story, whether it be about a bank robbery, dramatic rescue, or presidential candidate"s speech, must include the 5 w"s: what, who, where, when and why (sometimes cited as why/how).
