Statistical analysis used in Radiology, American Journal of Roentgenology, Chinese Journal of Radiology and Radiologic Practice: frequency and methods differences
邓明,王良,李亮,李拔森,蔡杰,谭芳琴,闵祥德,冯朝燕,Vivek K.Sah
DOI: https://doi.org/10.3760/cma.j.issn.1005-1201.2014.09.017
2014-01-01
Abstract:Objective To determine the types of statistical methods and their frequency used in major articles published in American and Chinese prestigious radiology journals,including Radiology,American Journal of Roentgenology (A JR),Chinese Journal of Radiology (C JR) and Radiologic Practice (RP).Methods One thousand and seventy-seven major articles published in Radiology,A JR,C JR and RP between September 2012 and August 2013 were retrospectively analyzed.The articles were exclusively original reports with objective,material and method,results and conclusion.All other articles,including case reports (fewer than 10 patients),review articles,pictorial essays,editorials,letters,and news items,were excluded from the study.The statistical method appearing in each article was assigned to one of eighteen categories (e.g.,basic t-test,x2 test,correlation and regression techniques,ANOVA,nonparametric statistics),and difference of the frequency between the American and the Chinese prestigious journals was assessed using Chi-square test.P＜0.05 denoted statistical significance.Results One thousand and seventy-seven major articles included C JR (n=210,19.498％),RP (n=278,25.813％),A JR (n=247,22.934％),and Radiology (n=342,31.755％).Two hundred and twenty-seven articles (21.077％) used no statisticalmethods or only descriptive statistics,including C JR (n=54,5.014％),RP (n=127,11.792％),A JR (n=31,2.878％),and Radiology (n=15,1.393％).Statistical methods were most frequently used in Radiology.There was statistical difference of frequency between the American (n=1391,72.789％) and the Chinese prestigious radiology journals (n=520,27.211％) (x2=197.836,P＜0.01).Basic statistical methods were more likely to be found in the articles of American journals (n=429,22.449％) than that in the Chinese journals (n=220,11.512％) (x2=138.197,P＜0.01).Correlation,regression analysis,ANOVA,and nonparametric statistics were more often used in the articles of American journals than that in the Chinese journals,which were n=245,12.820％ versus n=52,2.721％ (x2=137.549,P＜0.01) for correlation and regression analysis,n=98,5.128％versus n=31,1.622％ (x2=135.794,P＜0.01) for ANOVA,and n=216,11.302％ versus n=67,3.506％ (x2=138.197,P＜0.01) for nonparametric statistics.Conclusions The types of statistical methods and their frequency are quite different between the Chinese and the American prestigious radiology journals.Scientific research people should follow the scientific design and use ureasonable statistical methods to improve the quality of research papers.