You can't use two inequality constraints to simulate an equality constraint
in constrOptim().
-thomas
On Wed, 17 Mar 2010, Herb wrote:
>
> Hello at all,
>
> working with a dataset I try to optimize a non-linear function with
> constraint.
>
> test<-read.csv2("C:/Users/Herb/Desktop/Opti/NORM.csv")
>
> fkt<- function(x){
> a<-c(0)
> s<-c(0)
> #Minimizing square error
> for(j in 1:107){
> s<-(test[j,2] - (x[1] * test[j,3]) - (x[2] * test[j,4]) -
(x[3]*test[j,5]) -
> (x[4]*test[j,6]) - (x[5]*test[j,7]))^2
> a<- a+s}
> a<-as.double(a)
> return(a)}
>
> constrOptim(c(0,0,0.5,0,0.5), fkt, NULL, ui=rbind(c(1,1,1,1,1)), ci=c(1)
>
> It works perfekt in the previews way, adding a second constraint, to
> simulize the equal to 1 constraint, it doesn't work anymore:
>
> constrOptim(c(0,0,0.5,0,0.5), fkt, NULL,
> ui=rbind(c(1,1,1,1,1),c(-1,-1,-1,-1,-1)), ci=c(1,-1))
>
> the error message: initial value not feasible
>
> Thats the structure of the data file:
>
> test
> Date CPURNSA_2 SPTR_2 JNGWUS_2 CORP FNAR_2 CRY_2
> 1 28.02.2001 0.0040 -0.0912 -0.0043 -0.0104 0.0098 -0.0181
> 2 31.03.2001 0.0023 -0.0634 0.0088 -0.0519 0.0078 0.0013
> 3 30.04.2001 0.0040 0.0777 0.0202 0.0202 -0.0052 0.0213
> 4 31.05.2001 0.0045 0.0067 -0.0358 -0.0256 0.0102 0.0194
> 5 30.06.2001 0.0017 -0.0243 0.0032 -0.0165 0.0045 0.0508
> 6 31.07.2001 -0.0028 -0.0098 0.0072 -0.0139 0.0280 -0.0234
> 7 31.08.2001 0.0000 -0.0626 -0.0270 -0.0151 0.0108 0.0275
> 8 30.09.2001 0.0045 -0.0808 -0.0283 -0.0458 0.0028 -0.0476
> 9 31.10.2001 -0.0034 0.0191 -0.0034 -0.0254 0.0220 -0.0320
> .
> .
> .
> 107 31.12.2009 -0.0018 0.0193 0.0040 0.0216 -0.0100 0.0561
>
>
>
> May be anyone can see the problem or has an idea?
>
>
> Thanks in advance!
> Herb
> --
> View this message in context:
http://n4.nabble.com/constrOptim-error-initial-value-not-feasible-tp1596640p1596640.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> 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.
>
Thomas Lumley Assoc. Professor, Biostatistics
tlumley at u.washington.edu University of Washington, Seattle