[R] Error in UseMethod("predict")

[varin sacha]

Eric,

No problem. Let's see if somebody else has a solution. I have changed the smooth basis from P-splines ('ps') to thin plate ('tp'). It still does not work, but this time I get another error message.


##########################
#Data
y=c(34000,45000,19000,48900,65000,67000,78000,90000,51000,32000,54000,85000,38000,76345,87654,90990,78654,67894,56789,65432,18998,78987,67543,45678,76543,67876)
x=c(345,543,543,456,432,378,543,579,432,254,346,564,611,543,542,632,345,468,476,487,453,356,490,499,567,532)

Dataset=data.frame(y,x)

#Plot
plot(x,y)

#Robust GAM
library(robustgam)
true.family <- poisson()
fit=robustgam(x,y, sp=2424,family=true.family,smooth.basis='tp')
x.new <- seq(range(x)[1], range(x)[2])
robustfit.new <- pred.robustgam(fit, data.frame(X=x.new))$predict.values
lines(x.new, robustfit.new, col="green", lwd=2)

# To find the « sp » to include in the fit function here above
robustfit.gic<-robustgam.GIC.optim(x,y,family=true.family,p=3,c=1.6,show.msg=FALSE,smooth.basis="ps", method="L-BFGS-B")

## CROSS VALIDATION REPLICATIONS MSE ROBUST GAM

# Create a list to store the results
lst<-list()

# This statement does the repetitions (looping)
for(i in 1 :1000)
{

n=dim(Dataset)[1]
p=0.667
sam=sample(1 :n,floor(p*n),replace=FALSE)
Training =Dataset [sam,]
Testing = Dataset [-sam,]

fit18<-robustgam(x,y, sp=2424,family=true.family,smooth.basis='tp')

ypred=pred.robustgam(fit18,data.frame(X=Testing))
MSE = mean((y-ypred)^2)
MSE
lst[i]<-MSE
}
mean(unlist(lst))
####################################





Le dimanche 17 janvier 2021 à 15:02:45 UTC+1, Eric Berger <ericjberger using gmail.com> a écrit : 





Hi Sacha, 
I took a quick look. Sorry, I don't see immediately what is causing the problem.
Maybe someone else can help.



> Dear Eric,
> 
> Many thanks, I correct your 2 points and now I get another error message (Error in splineDesign(knots, x, ord, derivs, outer.ok = outer.ok, sparse = sparse) :
>   empty 'derivs'). 
> I have googleized and found some hints like (outer.ok=TRUE) but no one seems to work.
> 
> https://r.789695.n4.nabble.com/mgcv-gam-predict-problem-td3411006.html 
> 
> Any idea to make my code work would be appreciated.
> 
> Here below my new R code :
> 
> ##########################
> #Data
> y=c(34000,45000,19000,48900,65000,67000,78000,90000,51000,32000,54000,85000,38000,76345,87654,90990,78654,67894,56789,65432,18998,78987,67543,45678,76543,67876)
> x=c(345,543,543,456,432,378,543,579,432,254,346,564,611,543,542,632,345,468,476,487,453,356,490,499,567,532)
> 
> Dataset=data.frame(y,x)
> 
> #Plot
> plot(x,y)
> 
> #Robust GAM
> library(robustgam)
> true.family <- poisson()
> fit=robustgam(x,y, sp=2424,family=true.family,smooth.basis='ps',K=3)
> x.new <- seq(range(x)[1], range(x)[2])
> robustfit.new <- pred.robustgam(fit, data.frame(X=x.new))$predict.values
> lines(x.new, robustfit.new, col="green", lwd=2)
> 
> # To find the « sp » to include in the fit function here above
> robustfit.gic<-robustgam.GIC.optim(x,y,family=true.family,p=3,c=1.6,show.msg=FALSE,smooth.basis="ps", method="L-BFGS-B")
> 
> ## CROSS VALIDATION REPLICATIONS MSE ROBUST GAM
> install.packages("ISLR")
> library(ISLR)
> 
> # Create a list to store the results
> lst<-list()
> 
> # This statement does the repetitions (looping)
> for(i in 1 :1000)
> {
> 
> n=dim(Dataset)[1]
> p=0.667
> sam=sample(1 :n,floor(p*n),replace=FALSE)
> Training =Dataset [sam,]
> Testing = Dataset [-sam,] 
> 
> fit18<-robustgam(x,y, sp=4356,family=true.family,smooth.basis='ps',K=3)
> 
> ypred=pred.robustgam(fit18,data.frame(X=Testing))
> MSE = mean((y-ypred)^2)
> MSE
> lst[i]<-MSE
> }
> mean(unlist(lst))
> ####################################
> 
> 
> 
>  Le dimanche 17 janvier 2021 à 11:41:49 UTC+1, Eric Berger <ericjberger using gmail.com> a écrit : 
> 
> 
> Hi Sacha,
> I never used these packages before but I installed them and tried your code. I have a few observations that may help.
> 
> 1. the statement
>     ypred = predict(fit18,newdata=Testing)
>     is wrong. Checkout the help page (?robustgam)  which shows in the Examples section at the bottom to use something like
>     ypred = pred.robustgam( fit18, data.frame(X=Testing)
> 
> 2. your logic is wrong. You define the vectors x and y at the top. They should remain untouched during your program.
>     However in the loop you redefine y and then use the redefined y as an argument to robustgam() the next time through
>     the loop. This looks like a serious error.
> 
> HTH,
> Eric
> 
> 
> On Sun, Jan 17, 2021 at 12:20 PM varin sacha via R-help <r-help using r-project.org> wrote:
>> Dear R-experts,
>> 
>> Here below my reproducible R code. I get an error message (end of code) I can't solve.
>> Many thanks for your help.
>> 
>> ##########################
>> #Data
>> y=c(34000,45000,19000,48900,65000,67000,78000,90000,51000,32000,54000,85000,38000,76345,87654,90990,78654,67894,56789,65432,18998,78987,67543,45678,76543,67876)
>> x=c(345,543,543,456,432,378,543,579,432,254,346,564,611,543,542,632,345,468,476,487,453,356,490,499,567,532)
>> 
>> Dataset=data.frame(y,x)
>> 
>> #Plot
>> plot(x,y)
>> 
>> #Robust GAM
>> library(robustgam)
>> true.family <- poisson()
>> fit=robustgam(x,y, sp=4356,family=true.family,smooth.basis='ps',K=3)
>> x.new <- seq(range(x)[1], range(x)[2])
>> robustfit.new <- pred.robustgam(fit, data.frame(X=x.new))$predict.values
>> lines(x.new, robustfit.new, col="green", lwd=2)
>> 
>> # To find the « sp » to include in the fit function here above
>> robustfit.gic<-robustgam.GIC.optim(x,y,family=true.family,p=3,c=1.6,show.msg=FALSE,smooth.basis="tp", method="L-BFGS-B")
>> 
>> ## CROSS VALIDATION REPLICATIONS MSE ROBUST GAM
>> install.packages("ISLR")
>> library(ISLR)
>> 
>> # Create a list to store the results
>> lst<-list()
>> 
>> # This statement does the repetitions (looping)
>> for(i in 1 :1000)
>> {
>> 
>> n=dim(Dataset)[1]
>> p=0.667
>> sam=sample(1 :n,floor(p*n),replace=FALSE)
>> Training =Dataset [sam,]
>> Testing = Dataset [-sam,]
>> 
>> fit18<-robustgam(x,y, sp=4356,family=true.family,smooth.basis='ps',K=3)
>> 
>> ypred=predict(fit18,newdata=Testing)
>> y=Dataset[-sam,]$y
>> MSE = mean((y-ypred)^2)
>> MSE
>> lst[i]<-MSE
>> }
>> mean(unlist(lst))
>> ####################################
>>  
>> 
>> ______________________________________________
>> 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.
>> 
> 
>