Document Type : Review Article
Authors
1 Medical Informatics Research Center, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran
2 Center for Healthcare Data Modeling, Department of Biostatistics and Epidemiology, School of Public Health, Shahid Sadoughi University of Medical Sciences, Yazd, Iran
3 Department of Healthcare Services Management, School of Health Management & Information Sciences, Iran University of Medical Sciences, Tehran, Iran
Abstract
Introduction: Normality is a critical assumption for many statistical tests, particularly when sample sizes are small (less than 30 cases). Ensuring normality is essential for the validity of statistical analyses, making it imperative for researchers to verify this condition in their data.
Methods: To find pertinent published articles for this review, electronic resources like PubMed, Scopus, and Web of Science were methodically searched. To find pertinent publications, keywords like "normality tests," "medical data," and "statistical methods" were employed. Among the important tests covered are the Anderson-Darling test, Shapiro-Wilk test, Lilliefors corrected K-S test, and Kolmogorov-Smirnov test. The review gives a thorough overview of each method's applicability in various research situations by concentrating on its advantages and disadvantages.
Results: This analysis demonstrates that while there are several tests available to evaluate normality, different interpretations may cause misunderstandings. The Kolmogorov-Smirnov test may be favored for bigger data sets because of its wider applicability, but the Shapiro-Wilk test is typically taken into consideration for small sample sizes. The sensitivity of the Anderson-Darling test in identifying departures from normality is noteworthy. To select the best approach given the sample size and study environment, it is essential to comprehend the advantages and disadvantages of each test.
Conclusion: By clarifying the concepts and applications of normality tests, this study aimed to enhance researchers' understanding and effective use of these statistical tools in practice. The insights provided will help mitigate confusion surrounding normality assessments, ultimately improving the rigor and reliability of statistical analyses in research.
Highlights
Farzan Madadizadeh (Google Scholar) (PubMed)
Sajjad Bahariniya (Google Scholar) (PubMed)
Keywords
Main Subjects
)