Analysis Of Economic Growth Factor Based On Multiple Hypothesis Testing

To determine whether a variable in the model of economic growth is significant,it usually a p-value works,so we can judge whether the coefficient of variation is significant,and know whether or not the variable significantly affect economic growth.However,in economic growth multiple regression models,there are many variables.If followed by a separate inspection system variables at the significance level ?,to the whole test,the probability of error I will be ambitious larger than a,then to get some wrong conclusions.This article attempts to use multiple hypothesis testing techniques to solve the aforementioned problems,and use false discovery rate(FDR)control procedure to control the Type I error,then choose real factors determining economic growth.Through the simulation studies,using three given false discovery rate(FDR)algorithm derived value substantially higher than the nominal significance level,and with the correlation increasing(value from 0 to 0.5),it is reflected more obviously.As the correlation increases,multicollinearity between the explanatory variables is also more serious.Thus,classical single hypothesis testing becomes less secure.The FDR control process considers the dependency of the test statistics.Secondly,the use of false discovery three kinds FDR control algorithms derived false discovery rate than the traditional classic single hypothesis test weight test was much smaller rate.As empirical research,we use a data set(1997)of economist Sala-i-Martin,with the false discovery rate(FDR)process control,include BH algorithm,Storey algorithm and BKY algorithm,in order to select long-term impacting on countries' economic growth explanatory variables,to find that false discovery rate control is more accurate and objective.