Question

我无法强制rc样条线预测参考在函数内部时更改。我希望65岁为参考（yhat = 1，lower = 1，upper = 1）。当代码在“分析”功能之外时，它完美地工作。我怀疑当它在函数内时不会发生“更新”，即使dd $ limits确实变为参考65（参见输出）。可能是不同的环境？我试了几个小时来设置环境等，但遗憾的是没有成功。任何帮助将受到高度赞赏！

    library(Hmisc)
    library(survival)
    library(rms)
    library(cmprsk)

    bk.tst <- function(analysis) 
    {
    if (analysis=="test") {
        dt<-m
    }
    outcomes<-c("compos", "acs", "death", "vascdeath", "stroke", "majbleed")
    for (i in outcomes) {
     fvl<- dt$fvl
     age<-dt$age
     ndt <- data.frame(age, fvl)
     dd<-assign('dd', datadist(ndt), pos=1) 
     options(datadist='dd')
     tm<-paste("tt",i, sep="")
     SurvObj <- with(dt, Surv(eval(parse(text=tm)), eval(parse(text=i))==1))
     f<-cph(SurvObj ~ fvl*rcs(age,c(60,70,80)), type="Survival", method="exact", x=T, y=T)
     print(dd$limits) 
     dd$limits["Adjust to","age"] <- 65
     print(dd$limits)
     g <- update(f)
     ano<-anova(f)
     age.intr.rcsplines<-Predict(g, age=45:85, fvl, ref.zero=TRUE, fun=exp) 
     print(age.intr.rcsplines[20:22,])
     }
    }

    bk.tst("test")

#the output is:
                     age fvl
Low:effect      57.01027   0
Adjust to       64.62697   0
High:effect     71.69884   1
Low:prediction  39.48545   0
High:prediction 84.64122   1
Low             35.80287   0
High            92.15606   1
                     age fvl
Low:effect      57.01027   0
Adjust to       65.00000   0
High:effect     71.69884   1
Low:prediction  39.48545   0
High:prediction 84.64122   1
Low             35.80287   0
High            92.15606   1
   age fvl     yhat     lower    upper
20  64   0 0.986740 0.9599284 1.014300
21  65   0 1.008632 0.9929697 1.024541
22  66   0 1.034987 0.9799775 1.093084

# when the code is outside the function analysis, then the output is correct with 65 as a reference. That is what I want to happen also inside the function "analysis":
                     age fvl
Low:effect      57.01027   0
Adjust to       64.62697   0
High:effect     71.69884   1
Low:prediction  39.48545   0
High:prediction 84.64122   1
Low             35.80287   0
High            92.15606   1
                     age fvl
Low:effect      57.01027   0
Adjust to       65.00000   0
High:effect     71.69884   1
Low:prediction  39.48545   0
High:prediction 84.64122   1
Low             35.80287   0
High            92.15606   1
   age fvl      yhat     lower    upper
20  64   0 0.9782956 0.9369436 1.021473
21  65   0 1.0000000 1.0000000 1.000000
22  66   0 1.0261294 0.9868852 1.066934

Answer 1

最后，如下所示在函数bk.tst中定义环境有效！

bk.tst <- function(analysis, env = parent.frame()) 
{
if (analysis=="test") {
    dt<-m
}
outcomes<-c("compos", "acs", "death", "vascdeath", "stroke", "majbleed")
for (i in outcomes) {
 fvl<- dt$fvl
 age<-dt$age
 env$ndt <- data.frame(age, fvl)
 env$dd<-datadist(env$ndt) 
 options(datadist='dd')
 tm<-paste("tt",i, sep="")
 SurvObj <- with(dt, Surv(eval(parse(text=tm)), eval(parse(text=i))==1))
 env$f<-cph(SurvObj ~ fvl*rcs(age,c(60,70,80)), type="Survival", method="exact", x=T, y=T)
 print(env$dd$limits) 
 env$dd$limits["Adjust to","age"] <- 65
 print(env$dd$limits)
 g <- update(env$f)
 ano<-anova(g)
 age.intr.rcsplines<-Predict(g, age=45:85, fvl, ref.zero=TRUE, fun=exp) 
 print(age.intr.rcsplines[20:22,])
 }
}

bk.tst("test")

在r中的函数内改变参考值限制的三次样条

1 个答案: