Commonly Used Statistical Methods in This Journal

Abstract：Commonly used statistical methods: For categorical (nominal or ordinal) data, use the chi-square test and Ridit analysis. For comparison of interval variables between groups, use t-test, F-test, and rank sum test (suitable for interval variables that follow a skewed distribution). The application scope of the chi-square test: When the total sample size (n) > 40 and the expected frequency (T) > 5, use the basic formula of the chi-square test; if n > 40 and 1 < T < 5, use the corrected chi-square test; if n < 40 or T < 1, use the exact probability method of the chi-square test. T-test and F-test are suitable for interval variables that follow a normal distribution. The t-test is often used to compare the mean difference of paired design with the population mean of 0 (e.g., before and after treatment comparison), and the comparison of means between two small samples in a group design. For the comparison of means between two large samples in a group design, use the U-test. The F-test is a comprehensive comparison (it can only detect whether there is a statistically significant difference in the means of two or more groups, but cannot determine which two groups differ significantly). Multiple comparisons between the means of several samples can be performed using post-hoc tests (Post Hoc), commonly using LSD method, SNK method, Dunnett's method, Dunnett's T3 (for unequal variances), and Tamhane's T2 (for unequal variances).