Common statistical methods and software used in current dietetic research: An objective summary using text mining

摘要

Background: Statistical literacy is essential for evidence-based dietetic practice. Despite the fast-growing use of computationally-advanced statistical methods in recent decades, no data exist to identify which of these methods and software programs are commonly used in current dietetic research. Aim: To identify statistical methods and software frequently used in current dietetic research, with a view to informing statistics curricula and continuing education. Methods: A text mining approach using the bag-of-words method was applied to a random sample of articles obtained from all journals within the "Nutrition and Dietetics" subject category of SCImago Journal & Country Rank portal in 2018. A list of 229 statistical techniques and 19 software packages was developed to define search terms of interest. Statistical information from the methods section of included articles was extracted for data cleaning. Statistical analyses were conducted in R (Version 3.6.0) and Microsoft Excel (2016). Results: A total of 757 journal articles were included. Descriptive statistics were reported in more than 90% of articles. Introductory and intermediate inferential statistics were reported in more than 75% of articles, with chi-square test (30%), independent samples t-test (22%), ANOVA (21%) and logistic regression (16%) being most common. Multilevel models were the only advanced technique appearing in greater than 5% of articles (8.5%). SPSS was the most common software package (42%). Conclusions: These findings provide useful information for curriculum development and evaluation in dietetic education programs. They also inform educators of the necessary content to include in continuing education programs for dietitians and nutrition professionals.