Factor analysis results hinge on the researcher’s subjective procedural choices

Evidence

Uh, “rotate”? Yes, there are some esoteric concepts involved here. And while others who know a lot more about factor analysis than I do will disagree, my experience has been that you can get markedly different results depending on how you extract the factors, how you decide what number of factors to retain, how you rotate the factors, and (especially) how you interpret what you have left when you are done. (Beware of tables in journal articles that tell you how much an item loaded on only one factor, rather than on all of the factors.) — Bob Altemeyer, The Authoritarian Specter, 4016

Categorization

• XREF: Joins the vault’s methodological-caution thread about how quantitative findings can be researcher-driven, parallel to Milgrom’s point about institutions being shaped by their designer.