Clustering Standard Errors at the "Session" Level1

Session-specific features of a laboratory experiment, if those exist, do not disappear by clustering standard errors at the session level. Randomly ordering sessions, which is crucial to deal with sampling issues, cannot justify The experimental design should primarily determine In typical controlled experiment where subjects make choices in same environment repeatedly, participant level inherited from design, and could be larger (that is, statistical inference can more conservative) when clustered individual or decision-group than It implies that lead false-positive treatment effects it mistakenly chosen. A rule thumb using deviations introduced.