How Much Should We Trust the P-Value?


Ekici O.

12th International Statistics Days Conference ISDC2022, İzmir, Türkiye, 13 Ekim - 16 Kasım 2022, ss.33

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: İzmir
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.33
  • İstanbul Üniversitesi Adresli: Evet

Özet

page55image711895856

Notably, since 2013, the problems with the hypothesis testing approach and the use of p-value have been discussed intensively in the international literature. This study aims to activate the debates of the misuse of p-value in the national literature and thus contribute to raising awareness among the researchers. In the study, the discussions on the issue of significance testing are supported with authentic quantitative analyses. We interpret the critical aspects of the significance testing results with simulations. Simulation studies are generated to examine the types of error, the power of a test and the factors affecting the power. As one of the proposed solutions to the significance testing issues, whether reporting the effect size and the confidence intervals can be informative besides p-value is evaluated based on the simulation. The overall conclusion is that there is an obvious need to accept the limitations of statistical analysis and embrace the uncertainties in scientific research. The implications of this paper in the academic studies and statistics education done in both social and fundamental sciences will contribute to extending the correct use of p-value.