Psychological Repeatability Factors:Research Object Classification And Statistical Preconditions

There are many factors that affect the reproducibility of psychological research.This paper focuses on the matching of research objects and research paradigms.The premise of the statistical tools of psychological research assumes that these two factors have an impact on the reproducibility of psychological research.The empirical data are derived from the original data of 100 studies in "Estimating the reproducibility of psychological science".The study 1 was carried out using 87 repeated studies of data and results for analysis,normal fitting.The study found that(1)the study of cognitive psychology category relative to the study of social psychology category is more likely to be repeated success,cognitive psychology in the field of success accounted for 52.4%(42 studies),social psychology in the field of repeated success accounted for 22.2%(45studies);(2)the vast majority of statistical means of the data assumptions are based on the normal distribution basis,through the normal fitting analysis found that,87 studies did not meet all the psychological characteristics of the hypothesis,87 studies were totally fited with 650 times,which accounted for 48.3%of the normal distribution(314 conform to the normal distribution);(3)After removing the maximum value,and the proportion of cognitive psychology and social psychology data is not very different from the normality.The 51.3%study of social psychology did not correspond to the proportion of normal distribution.The 57.1%study of cognitive psychology did not conform to the normal distribution,Study 2 uses the empirical results of 97 studies to explore the impact of statistical assumptions on statistical studies in psychology on repetitive studies.The study found that(1)the use of T test as a statistical tool to study the success rate is much higher than the use of variance analysis as a statistical tool for the study,the former repeat success rate of 52.2%(12/23),the latter repeat success rate of 32.8%(20/61),that is,in the psychological study of T-test robustness stronger than the variance analysis;(2)in the use of variance analysis as a statistical means of research,whether to meet the normative assumptions and variance homogeneity of the two assumptions seem right The success rate of repetition is 27.7%,and the success rate of the study does not accord with the normality hypothesis.The success rate is 30.7%.In the data consistent with the variance homogeneity hypothesis,the research is repeated successfully.Accounting for 15.6%;in the data that does not conform to the variance homogeneity hypothesis,the proportion of the study success is 52.0%.Based on the results of Study 1 and Study 2,this paper argues that the rough classification of the study according to the article is not appropriate,try to use machine learning based on the text of the article analysis,the results of three studies show that(1)the topic modeling fits better than clustering;(2)the theme of the theme extracted and the results of the block is still more consistent with the cognitive-social psychology division,did not follow the success of the study to distinguish between success or failure.