Validating arguments for observational instruments: Attending to multiple sources of variation.