The first step to good clinical research: statistical thinking

The Korean Journal of Anesthesiology, the official journal of the Korean Society of Anesthesiologists, recently introduced the Statistical Round as a new publishing category for the year 2015. The Aims and Scope section of the journal explains that the Statistical Round category provides educational material to help the reader understand the designs of experimental and statistical analyses used in anesthesiology investigations. A Statistical Round is a narrative review of the application of various topics ranging from contemporary quantitative sciences to issues of concern to anesthesia researchers. A Statistical Round provides a focused discussion of one or more unique or interesting statistical analysis methods that were previously published in this journal or express the general policies or opinions of the Statistical Round Board. Statistical Round articles are provided by members of the Statistical Round Board or are invited from authors and then are reviewed by the Statistical Editor. All of the Statistical Round articles are published in both English and Korean for the convenience of Korean readers; the Korean version is present only on the journal's web page. Statistics is the practice or science of collecting and analyzing numerical data in large quantities, especially for the purpose of inferring proportions in a whole population from those in a representative sample. Statistical Round articles ultimately provide guidance for performing statistical analysis, which could be very helpful for both authors and reviewers. A good example of these goals can be found in other medical science journals (e.g., Brit-ish Medical Journal, Anesthesia & Analgesia) [1]. Many data obtained through research are questioned. How great should the difference in means between two samples be to consider the presence of a meaningful difference between the two populations? If the sample mean does not represent the average of the population, it is correct to analyze data using the t-test in such cases? What nonparametric methods correspond to one-factor analysis of variance (ANOVA)? When the null hypothesis is rejected after one-factor ANOVA, how can we determine whether all means are different between each groups? What would the result be of ignoring the nature of the repeated-measure data if an independent t-test is used to compare two groups? Can we only use continuous or categorical data to measure the level of the dependent variable in multiple regression analysis? Can we measure the levels of independent variables using both continuous and categorical data? Are any problems evident between the relevant …