Evaluation of Educational Scientific Articles with the Purpose of Identification Common Statistical Errors

The goal of this research is to identify the statistical errors in the articles published in the field of Educational Sciences. For this purpose, by using a two-stage clustering sampling method, 92 articles were selected out of 12 issues of the journals in Educational Sciences studied. The research design was based on document analysis and the method for analyzing the data was according to multiple comparisons by the use of error bars. The results showed that 39 articles out of 92 ones had at least one statistical error, and among this article, there were 49 common error including, error types listed based on frequency of occurrence: "Fitting the regression model without examining the assumptions of the residuals, Use of independent T-test without examining the homogeneity of variances (levene's test), false fitting of the structural equations model, Biased sampling method, Using the analysis of variance instead of T-test, Using Friedman's nonparametric test without considering the distribution of data, Mismatch results of the statistical analysis, Using the independent T-test instead of paired one, Interpreting the results of Levene's test instead of the results of T-test, Performing the follow-up test without performing the analysis of variance, and finally Analysis of statistical significance just based on the results of descriptive statistics". The results also showed that the highest frequency of errors is related to the extracted articles out of master's theses, organizational affliction related to Islamic Azad University, academic degree of the instructor, non-specialists in Educational Sciences, journals with a history of publication below 20 years and article with 3 authors.