[R] 95% bootstrap CIs

varin sacha v@r|n@@ch@ @end|ng |rom y@hoo@|r
Thu Sep 26 23:13:56 CEST 2019


Dear Rui,

Excellent ! Many thanks.








Le mercredi 25 septembre 2019 à 18:50:09 UTC+2, Rui Barradas <ruipbarradas using sapo.pt> a écrit : 





Hello,

In your reproducible example you forget to define 'data'.
You should also

set.seed(<some_int_number>)


The following works.


data <- data.frame(a, x, z, y_obs)
boot.ci.type <- c("norm","basic", "perc")

mse_gam <- function(data,i) {
  boot.gam <- gam(y_obs~s(x)+s(z)+s(a),data=data[i,])
  mean(boot.gam$residuals^2)
}

bootResults_gam <-boot(data=data, statistic=mse_gam, R=1000)
boot.ci(bootResults_gam, type = boot.ci.type)


mse <- function(data,i) {
  boot.earth <- earth((y_obs~x+z+a),data=data[i,])
  mean(boot.earth$residuals^2)
}

bootResults <- boot(data=data, statistic=mse, R=1000)
boot.ci(bootResults, type = boot.ci.type)



Hope this helps,

Rui Barradas

Às 13:43 de 25/09/19, varin sacha via R-help escreveu:
> Dear R-experts,
> 
> Below the reproducible example. I have tried to write a function that returns the statistic of interest (MSE in my case). I have run boot( ) where the function is included in the statistic argument. I have run boot.ci with the result from boot( ). I guess the error comes from the data : bootResults <- boot(data=?????,statistic=mse, R=1000)
> Many thanks for your help.
> 
> ##################################################
> library(mgcv)
> 
> library(earth)
> 
> library(boot)
> 
>  
> n<-2000
> 
> x <-runif(n, 0, 5)
> 
> z <- rnorm(n, 2, 3)
> 
> a <- runif(n, 0, 5)
> 
> 
> y_model<- 0.1*x^3 - 0.5 * z^2 - a + 10
> 
> y_obs<-rnorm(n, y_model, 0.1)
> 
> gam_model<- gam(y_obs~s(x)+s(z)+s(a))
> 
> mars_model<-earth(y_obs~x+z+a)
> 
>  
> MSE_GAM<-mean((gam_model$fitted.values - y_model)^2)
> 
> MSE_MARS<-mean((mars_model$fitted.values - y_model)^2)
> 
>  
> MSE_GAM
> 
> MSE_MARS
> 
>  
> 
> mse <- function(data,i) {
> 
> boot.gam <- gam(y_obs~s(x)+s(z)+s(a),data=data[i,])
> 
> return(mean(boot.gam$residuals^2))
> 
> }
> 
> bootResults <-boot(data=data,statistic=mse,R=1000)
> 
>  
> 
> mse <- function(data,i) {
> 
> boot.earth <- earth((y_obs~x+z+a),data=data[i,])
> 
> return(mean(boot.earth$residuals^2))
> 
> }
> 
> bootResults <-boot(data=data,statistic=mse,R=1000)
> ##################################################
> 
>  
> 
> 
> 
> 
> 
> 
> 
> 
> 
> Le lundi 23 septembre 2019 à 21:42:56 UTC+2, varin sacha via R-help <r-help using r-project.org> a écrit :
> 
> 
> 
> 
> 
> Dear R-Experts,
> 
> Here is my reproducible R code to get the Mean squared error of GAM and MARS after I = 50 iterations/replications.
> If I want to get the 95% bootstrap CIs around the MSE of GAM and around the MSE of MARS, how can I complete/modify my R code ?
> 
> Many thanks for your precious help.
> 
> ##################
> 
> library(mgcv)
> library(earth)
> my.experiment <- function() {
> n<-500
> x <-runif(n, 0, 5)
> z <- rnorm(n, 2, 3)
> a <- runif(n, 0, 5)
> y_model <- 0.1*x^3 - 0.5*z^2 - a + x*z + x*a + 3*x*a*z + 10
> y_obs <- y_model +c( rnorm(n*0.97, 0, 0.1), rnorm(n*0.03, 0, 0.5) )
> gam_model<- gam(y_obs~s(x)+s(z)+s(a))
> mars_model<-earth(y_obs~x+z+a)
> MSE_GAM<-mean((gam_model$fitted.values - y_model)^2)
> MSE_MARS<-mean((mars_model$fitted.values - y_model)^2)
> return( c(MSE_GAM, MSE_MARS) )
> }
> my.data = t(replicate( 50, my.experiment() ))
> colnames(my.data) <- c("MSE_GAM", "MSE_MARS")
> summary(my.data)
> 
> ##################
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 



More information about the R-help mailing list