Dear Peter Dalgaard, Really appreciated, but my code does not work. There is still a problem ! Here below the reproducible example with 20 variables
library(mgcv) library(earth) n<-2000 x<-runif(n, 0, 5) z <- rnorm(n, 2, 3) a <- runif(n, 0, 5) b <- rnorm(n, 2, 3) c <- runif(n, 0, 5) d <- rnorm(n, 2, 3) e <- runif(n, 0, 5) f <- rnorm(n, 2, 3) g <- runif(n, 0, 5) h <- rnorm(n, 2, 3) i <-runif(n, 0, 5) j <-rnorm(n, 2, 3) k <-runif(n, 0, 5) l <-rnorm(n, 2, 3) m <-runif(n, 0, 5) n <-rnorm(n, 2, 3) o <-runif(n, 0, 5) p <-rnorm(n, 2, 3) q <-runif(n, 0, 5) r <-rnorm(n, 2, 3) y_model<- 0.1*x^3 - 0.5 * z^2 - a^2 + b^2 + 2*c + 3*d - 4*e + 3.5*f - 4.5*g+ 2.5*h + 5.5*i^2 -1.5*j - 6*k + l + 2*m + n + 3*o - 4.5*p + q - r + 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)+s(b)+s(c)+s(d)+s(e)+s(f)+s(g)+s(h)+s(i)+s(j)+s(k)+s(l)+s(m)+s(n)+s(o)+s(p)+s(q)+s(r)) mars_model<-earth(y_obs~x+z+a+b+c+d+e+f+g+h+i+j+k+l+m+n+o+p+q+r) MSE_GAM<-mean((gam_model$fitted.values - y_model)^2) MSE_MARS<-mean((mars_model$fitted.values - y_model)^2) MSE_GAM MSE_MARS Le mercredi 18 septembre 2019 à 11:04:26 UTC+2, peter dalgaard <pda...@gmail.com> a écrit : Um, I think not... The mean of the last 200 observation won't line up with the x and z. Possibly, if what you want is the last 200 obs to have a different variance, y_obs <- y_model + c(rnorm(0.9 * n, 0, 0.1), rnorm(0.1 * n, 0, 0.5)) or y_obs <- rnorm(n, y_model, rep(c(0.1, 0.5), c(.9 * n, .1 * n))) -pd > On 17 Sep 2019, at 22:27 , David Winsemius <dwinsem...@comcast.net> wrote: > > > On 9/17/19 12:48 PM, varin sacha via R-help wrote: >> Dear R-helpers, >> >> Doing dput(x) and dput(y_obs), the 2 vectors are not the same length (1800 >> for y_obs and 2000 for x) >> How can I solve the problem ? >> >> Here is the reproducible R code >> >> # # # # # # # # # # >> library(mgcv) >> library(earth) >> >> n<-2000 >> x<-runif(n, 0, 5) >> y_model<- 0.1*x^3 - 0.5 * x^2 - x + 10 >> # y_obs<-rnorm(n*0.9, y_model, 0.1)+rnorm(n*0.1, y_model, 0.5) # maybe not >> exactly your goal? > > > You didn't lay out any goals for analysis, so let me guess what was intended: > > > I suspect that you were hoping to model a mixture composed of 90% from one > distribution and 10% from another. If I'm right about that guess then you > would instead wat to join the samples from each distribution: > > y_obs<-c( rnorm(n*0.9, y_model, 0.1), rnorm(n*0.1, y_model, 0.5) ) > > -- > > David > > >> gam_model<- gam(y_obs~s(x)) >> mars_model<- earth(y_obs~x) >> MSE_GAM<-mean((gam_model$fitted.values - y_model)^2) >> MSE_MARS<-mean((mars_model$fitted.values - y_model)^2) >> MSE_GAM >> MSE_MARS >> # # # # # # # # # # # # # # # # >> >> ______________________________________________ >> R-help@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@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. -- Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Office: A 4.23 Email: pd....@cbs.dk Priv: pda...@gmail.com ______________________________________________ R-help@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@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.
