rm(list=ls()) library("BMS") load("replication_JIMF.RData") # X.data contains the explanatory variables, y.data the different EMP measures; X.int is the extended set of explanatory variables containing also linear interaction terms iterNr=2e05;burnNr=3e05 # the results in the article are based on 1 million burn-ins and 3 million iterations (posterior draws) resList=resList.int=list() resList$emp.max=try(bms(cbind(y.data[,"EMPu_max",drop=F],X.data),iter=iterNr,burn=burnNr,g="hyperUIP",mprior="random")) resList$emp.max0006=try(bms(cbind(y.data[,"EMPu_max.0006",drop=F],X.data),iter=iterNr,burn=burnNr,g="hyperUIP",mprior="random")) resList$emp.ptp=try(bms(cbind(y.data[,"EMPu_PtT",drop=F],X.data),iter=iterNr,burn=burnNr,g="hyperUIP",mprior="random")) load("bmsInt.RData") # this is an add-on to BMS in order to use the strong / weak heredity prior a la H. Chipman (1996) for linear interaction terms; the names of the lin. interaction terms have to be separated by # (e.g., a#b) resList.int$EMP.max<-bms.int(cbind(y.data[,"EMPu_max",drop=F],X.int),iter=iterNr,burn=burnNr,g="hyperUIP",mprior="random",heredP=c(0,0,1)) resList.int$EMP.max0006<-bms.int(cbind(y.data[,"EMPu_max.0006",drop=F],X.int),iter=iterNr,burn=burnNr,g="hyperUIP",mprior="random",heredP=c(0,0,1)) resList.int$EMP.ptp<-bms.int(cbind(y.data[,"EMPu_PtT",drop=F],X.int),iter=iterNr,burn=burnNr,g="hyperUIP",mprior="random",heredP=c(0,0,1),start.value=0)