POL SCI 137A Chapter Notes - Chapter 7: Subgroup Analysis, Statistical Hypothesis Testing
Document Summary
Fishing: researchers intentionally and unintentionally selecting models that yield positive findings, leading to an unreliable body of published research the realization of the conclusion. Model is fished when the decision to report the model depends on. Problem: bias induced by selection of what results get reported. This problem is usually discussed in terms of reporting and publication biases, but also enters at other stages (analysis, discussions, time of submission - file drawer bias, in attention paid to results by others) Enabled because social scientists generally have substantial latitude in. Preregistration to combat fishing by leading journals - prior identified selecting measures and models for hypothesis testing. Also quality of registration in medical research varies substantially. Even when standards of medical registries are met, info about modeling decisions that are routine for ps analysis is absent. Registration procedures in medicine still provide researchers with substantial leeway in practice.