Assessment Of Statistical Power In Contemporary Accounting Information Systems Research

The purpose of this study is to provide a current, representative assessment of statistical power in accounting information systems (AIS) research. This study empirically investigates whether the statistical power of extant AIS research has been strong enough to detect important relationships that may exist. A power analysis of 45 articles from the most recent, complete five years (1996-2000) of Journal of Information Systems and Journal of Management Information Systems shows that, on the average, 56 percent of empirical studies do not have high power levels. This suggests that, on average across all effect sizes, more than half the time AIS researchers risk not being able to detect significant effects when, in fact, they exist. This risk increases greatly as the effect size decreases. Current findings suggest the need for more statistical power planning in AIS research designs. Statistical power is important to AIS research because it increases the probability of making correct decisions about empirical studies. Without adequate statistical power, AIS research may fail to identify statistically significant results and viable research streams might be abandoned prematurely. Statistical power will also become increasingly important as empirical studies in AIS study relatively smaller effects.