Résumé
Objective:It introduced the application of bayesian network in real-world data and the implementation of Tetrad software, so as to provide methodological application reference for medical researchers.Methods:Real-world data for diabetes diagnosis had been taken as an example. UCI machine learning database Pima-Indian-diabetes data set was used as case data, Bayesian network ratiocination was carried out according to the construction order of bayesian network of Tetrad software.Results:When the screening age was over 30 years old, impaired glucose tolerance, abnormal serum insulin two hours after meal, body weight index and family genetic index were large, the probability of diabetes onset from 34.99% to 83.33%.Conclusions:The bayesian network realized by Tetrad software provides a powerful tool for causal inference of real-world data, and promoted the objective and quantitative interpretation of the dependence between variables.