Experimental studies determine causal relationships between constructs by manipulating independent variables and analysing their effects on dependent variables. What should we do with samples that do not pass the manipulation checks? Can we simply exclude them?