That is, the random assignment of treatments to experimental units simulates the effect of independence and allows the researcher to conduct tests and estimation procedures as if the observed responses were randomly selected from an existing population.