A Probabilistic Filtering Approach to Non‐Effortful Responding

Common response-time-based approaches for non-effortful response behavior (NRB) in educational achievement tests filter responses that are associated with times below some threshold. These are, however, limited they require a binary decision on whether is classified as stemming from NRB; thus ignoring potential classification uncertainty resulting parameter estimates. We developed probabilistic filtering procedure overcomes this limitation. The rooted the principles of multiple imputation. Instead creating plausible replacements missing data, data sets created represent filtered data. propose two different to models, originating research traditions and conceptualizations identification NRB. first approach uses Gaussian mixture modeling identify time subcomponent Plausible based examinees' posterior probabilities belonging NRB subcomponent. second defines range thresholds creates by drawing defined range. illustrate workings proposed well differences between models both simulated empirical PISA 2018.