The Use of Effect Size in Veterinary Medicine: An Alternative Approach to p Value

Abstract

Effect size is a statistical index that measures the magnitude of the effect generated by the variable of interest in a study, in a sense, reflecting the practical or clinical value of the study in addition to the statistical results. In recent years, it has become preferable to report the effect size expressing practical significance in addition to the statistical significance expressed by the p-value in hypothesis tests in scientific research, and even it has been required by some scientific journals. By reporting the effect size, it is possible to use it in statistical power analysis, to compare the results of the studies, and to determine the amount of the effect in the study. In this study, by mentioning the concept of effect size, the main effect size indices used according to research types are introduced. In addition, the calculation methods of the effect size indices commonly used for continuous and categorical outcome variables were given and interpreted with scenarios from the field of veterinary medicine. In conclusion, in order to be able to interpret the results of a study in clinical or practical terms, to present the analyzed data in more detail than the p-value, and to ensure its use in power analysis, it was suggested that researchers report effect size in their studies. Etki büyüklüğünün veteriner hekimliği alanında kullanımı ÖZ Etki büyüklüğü, bir çalışmada ilgilenilen değişkenin meydana getirdiği etkinin büyüklüğünü ölçen, bir anlamda çalışmanın istatistiksel sonuçlarına ek olarak, pratik veya klinik anlamdaki değerini de yansıtan istatistiksel bir indekstir. Son yıllarda bilimsel araştırmalarda hipotez testlerinde p değeri ile ifade edilen istatistiksel anlamlılığa ek olarak pratik anlamlılığı ifade eden etki büyüklüğünün de raporlanması tercih edilir hale gelmiş, hatta bazı bilimsel dergiler tarafından zorunlu kılınmıştır. Etki büyüklüğünün raporlanması ile istatistiksel güç analizinde yararlanılması, çalışmaların sonuçlarının karşılaştırılması ve çalışmada belirlenen etkinin miktarının belirlenmesi mümkün olmaktadır. Bu çalışmada etki büyüklüğü kavramından bahsedilerek, araştırma türlerine göre kullanılan başlıca etki büyüklüğü indeksleri tanıtılmıştır. Ayrıca, sürekli ve kategorik sonuç değişkenler için yaygın olarak kullanılan etki büyüklüğü indekslerinin hesaplama yöntemleri verilerek, veteriner hekimliği alanından senaryolar ile örneklendirilmiş ve yorumlanmıştır. Sonuç olarak, klinik veya pratik anlamda çalışma sonuçlarını yorumlayabilmek, incelenen veriyi p değerinden daha ayrıntılı şekilde sunabilmek, güç analizlerinde kullanımını sağlamak amacıyla, araştırmacıların çalışmalarında etki büyüklüğü raporlaması önerilmiştir. 1Department of Biostatistic, Faculty of Veterinary Medicine, Mustafa Kemal University, Hatay/TURKEY MAE Vet Fak Derg, 6 (1): 2833, 2021 DOI: 10.24880/maeuvfd.823493 Bu makaleye atıfta bulunmak için/ To cite this article: Ambarcıoğlu P. The use of effect size in veterinary medicine. MAE Vet Fak Derg. 2021; 6 (1) :28-33. INTRODUCTION Statistical significance tests have a history dating back to the 1700s. It was first used by the Scottish physician John Arbuthnot in evaluating the birth rate in London according to the sex of the newborn babies, but its use was not widespread until the 1900s (1). The use of hypothesis tests has become widespread with the development of Karl Pearson’s Chi-Square Goodness of Fit Test in 1900, William S. Gossett’s Student t-Test in 1908, and Ronald Fisher’s Analysis of Variance (ANOVA) in 1918 (2,3,4). It became popular after Fisher published his first book in 1925, Statistical Methods for Research Workers, and then The Design of Experiments in 1935 (1,5,6). Although the decision-making process based on observation data in scientific research is actually a complex process, the fact that the deterministic algorithm of hypothesis testing approach has reduced this process to a dichotomous form of accepting and rejecting the hypothesis. For this reason, the hypothesis testing approach has spread rapidly as an easy-to-use process for researchers (7). However, over time, as researchers perceived this process only as a tool used to reach statistical significance, it has become difficult to reach scientific generalizations, and consequently, it received serious criticism (8,9,10). Criticisms of the hypothesis testing approach argue that the effect size should be used to express practical or clinical significance, in addition to the statistical significance expressed by hypothesis testing. Indeed, organizations such as the American Educational Research Association (AERA) and the American Review