Stimulus as a random factor in analysis of variance: Increasing the generalizability of findings

All too often inadequate experimental designs and/or faulty statistical analyses preclude generalization of findings. The authors of this paper discuss the need, in designing studies having educational variables, to give more attention to recognizing the universe of generalization and to incorporating into the design a variable that represents this universe. They further recommend that data analytic techniques, in turn, treat the levels of this variable as randomly selected from the universe.