| In view of universal misuse of medical statistics, especially in medical journals, this study aims to explore a new teaching method different from the traditional one. Medical papers published in journals in the last ten years have been reviewed, the cases with typical statistical errors categorized, offered in the original form, analyzed according to the error type, and given corresponding countermeasures.The typical statistical errors in medical papers are as follows:1. In the research design, common errors are absence of a perfect research design plan, violation of the principles of randomization, control, replication and balance, use of the single-factor design instead of the multi-factor design, and failure to collect data in accordance with the research design.2. In the data's expression and description, common errors are reversion of the column title and the row title when drawing statistical tables, mixture of different contents in the same table, vague expression of table's data, inconsistency of statistical plot and the type of data, failure to abide by mathematical principles when drawing the scales on coordinate axis, use of 'MEAND(SEM)' to describe all quantitative data, too small denominator when using relative numbers, confusions of ratios and rates.3. In the statistical analysis of quantitative data, common errors are use of statistical methods of qualitative data because of misunderstanding of data's nature, use of the parametric test without discerning the precondition, using of / test or ANOVA for one-factor design to process all kinds of quantitative data regardless of design types.4. In the statistical analysis of qualitative data, common errors are misuse of / test to analyze qualitative data, use of chi-square test to analyze all kinds of qualitative data without determining the nature of qualitative variables, disagreement of statistical methods with research aims.5. In the statistical analysis of correlation and regression, common errors are confusion of correlation and regression, use of type I regression data for correlation, analysis of the data without the linear trend in the scatter plot, interpretation of the results of correlation coefficient only based on hypothesis testing, interpretation of thecorrelation as causality.6. In the statistical analysis of multi variable data, common errors are use of one-variable analysis instead of multi variable analysis, use of one-factor analysis instead of multi factor analysis, inability of the variables which make sense only accompanied by other variables to enter the equation during multi variable linear regression analysis because of the wrong methods, inability to choose the correct methods based on various analysis aims.7. In the statistical resulting explanation, common errors are making direct conclusions without employing statistics or explaining the methods to be used, presentation of P values without specifying the test used, inconsistency of statistical methods and results of explanation, low statistical power, misunderstanding the meaning of P value, and inferring causality only based on association.To prevent the misuse of statistics in medical journals, the corresponding governmental supervision and administration department should make policies and enforce supervision. By presenting original medical papers and identifying statistical errors like 'actual combat', practical scientific staff can understand statistical concepts clearly, apply correct statistical strategies, solve actual problems using statistics flexibly, reduce the chance of misusing statistics, for the purpose of exerting the power of statistics and improve the scientific validity and rigor of medical research.In this study, most kinds of typical statistical errors in medical papers are collected, and a computer-aided instruction system is built, the contents for discussion and corresponding point knowledge are displayed clearly, the advantages of hyper-text are fully presented, a...
|
PDF Full Text Request