Main content
Good tests replicate
Date created: | Last Updated:
: DOI | ARK
Creating DOI. Please wait...
Category: Project
Description: What does using a good test mean, and why should psychologists care? From a purely statistical perspective, good tests make practical assumptions about the kinds of data they are applied to, and they perform well when those assumptions are met (e.g., they maintain a 5% false positive error rate). Replication depends in part on whether psychologists are mindful about these assumptions—and the consequences of violating them—when conducting and interpreting test results. Often times, these assumptions are carelessly or unwittingly violated. In this presentation, I will demonstrate how “significant” findings can dramatically mislead and thus also fail to replicate when assumptions are not met. From the perspective of psychological theory, good tests stem from good study design, and their results clearly (dis)confirm specific predictions. Here, replication depends on unambiguous research questions and specific predictions. However, many times tests are used that do not answer the question at hand; tests can be better tailored to specific questions, and tests that better suit research questions are easier to interpret in both original and replication studies. In this presentation, I will give specific examples of how psychologists can increase replication not only by ensuring that their tests meet statistical assumptions but also by ensuring that their research questions easily lend themselves to specific tests.