Important Issues in Statistical Testing and Recommended Improvements in Accounting Research

Econometrics

ThomasR. Dyckman¹  StephenA. Zeff² 

[1] Accounting Department, Cornell University, Ithaca, NY 14850, USA;Accounting Department, Rice University, Houston, TX 77005, USA;

关键词: model specification;    model testing;    reporting results (p-values);    replications;   

DOI  :  10.3390/econometrics7020018

来源: DOAJ

【 摘 要 】

A great deal of the accounting research published in recent years has involved statistical tests. Our paper proposes improvements to both the quality and execution of such research. We address the following limitations in current research that appear to us to be ignored or used inappropriately: (1) unaddressed situational effects resulting from model limitations and what has been referred to as “data carpentry,” (2) limitations and alternatives to winsorizing, (3) necessary improvements to relying on a study’s calculated “p-values” instead of on the economic or behavioral importance of the results, and (4) the information loss incurred by under-valuing what can and cannot be learned from replications.

【 授权许可】

Unknown