Masic Izet, Jankovic Slobodan M, Begic Edin
Academy of Medical Sciences of Bosnia and Herzegovina, Sarajevo, Sarajevo, Bosnia and Herzegovina.
Department of Pharmacology, University of Kragujevac, Faculty of Medical Sciences, Kragujevac, Serbia.
Stud Health Technol Inform. 2019 Jul 4;262:105-109. doi: 10.3233/SHTI190028.
Correct choice and administration of a statistical test are absolutely essential for meaningful interpretation of research data, yet mistakes are still frequent and could be easily found in published scientific papers or PhD theses. The aim of this study was to analyze mistakes made by PhD students in statistical analysis of data collected during research within the framework of their thesis. PhD students frequently use Excel and SPSS for data processing, while SAS, Stata and R are also available. The study was designed as cross-sectional analysis of random sample (n=15) of PhD theses in pre-approval stage at Faculty of Medical Sciences, University of Kragujevac, Serbia. In total 14 (93%) theses had at least one mistake. The most frequent mistakes were as the following: insufficient statistical power due to small sample size, inappropriate presentation of results at tables and graphs, andinappropriate choice of statistical tests. In order to improve the situation, training courses in statistics during PhD studies should be re-evaluated and improved in regard to relevance, delivery methods and motivating potential, and mentors should invest more effort to review the data and guide students through statistical analysis.
正确选择和应用统计检验对于有意义地解释研究数据绝对至关重要，但错误仍然屡见不鲜，在已发表的科学论文或博士论文中很容易发现这些错误。本研究的目的是分析博士生在其论文框架内研究过程中收集的数据进行统计分析时所犯的错误。博士生经常使用Excel和SPSS进行数据处理，同时也可使用SAS、Stata和R。该研究设计为对塞尔维亚克拉古耶瓦茨大学医学系处于预批准阶段的博士论文随机样本（n = 15）进行横断面分析。总共14篇（93%）论文至少有一个错误。最常见的错误如下：样本量小导致统计效力不足、表格和图表中结果呈现不当以及统计检验选择不当。为改善这种情况，应重新评估并改进博士学习期间的统计学培训课程，包括相关性、授课方式和激励潜力等方面，导师应投入更多精力审查数据并指导学生进行统计分析。
