Regression artifacts are a source of mistaken causal inference in inferences based on time-series data and from longitudinal studies. These artifacts are illustrated, and it is noted that their magnitude is computable (and distinguishable from genuine effects) if the autocorrelation patterns for various lags is known. (SLD)

In most cases, questions of cause and effect cannot be answered by social science methods. However, it is argued that evaluation should not be freed from dealing with missing causal links. It is possible to predict whether a causal question is likely to be measured or assessed. (SLD)