# [R] bootstrap and nls

Spencer Graves spencer.graves at pdf.com
Sun Jul 18 15:45:53 CEST 2004

```      1.  Have you studied the documentation, including working the
examples, for "nls" and "boot"?  I don't see "data" as an argument to
"nls", as I'm used to, and your use of "i" also seems disconcerting to
me.  Something like the following seems to me to be closer to the
standard syntax and therefore safer:

nls(SolA~a/(1+b*Tps)^c,start=c(a=31,b=0.5,c=0.6), data=data[i,])

This may not help with this problem, but it might help you with others
-- possibly even in understanding how R is structured more generally.

2.  If this error occurs in only a very few cases and you can
afford to ignore such cases, then you could use "try".  I didn't
remember exactly the way to do this, so I performed the following
experiment:

> Er <- try(if(NA)2)
Error in if (NA) 2 : missing value where TRUE/FALSE needed
> class(Er)
 "try-error"

This leads me to suggest the following:

dif.param<-function(data,i){
RaiesLossA.nls<-try(nls(SolA[i]~a/(1+b*Tps[i])^c,start=c(a=31,b=0.5,c=0.6)))
RaiesLossB.nls<-try(nls(Solb[i]~a/(1+b*Tps[i])^c,start=c(a=33,b=1.4,c=0.5)))
if((class(RaiesLossA.nls)=="try-error") |
(class(RaiesLossB.nls)=="try-error"))
return("whatever you want in this case")
else RaiesLossA.nls\$m\$getPars()-RaiesLossB.nls\$m\$getPars()
}

#  UNTRIED

3.  What is the nature of your data and the physical context?  Can
the deviations from this model reasonably be considered to be
independent normal with constant variance?  If yes, then you have a
reasonable model.  Alternatively, if all numbers are positive, have you
considered the following:

log(SolA) = ln.a + c*log(1+b*Tps)

If the deviations from this model seem close to being normally
distributed, then you might try using this form with method="plinear".
If you do this, you only need starting values for "b".  Whether you use
"plinear" or not, have you tried getting starting values as follows:
Recall that log(1+x) = x+x^2/2+O(x^2).  If b*Tps is usually small, then
the following might work to produce reasonable starting values:

fit0 <- lm(SolA~Tps + I(Tps^2), data=data[i,])

Then coef(fit0) estimates log(a), c*b and c*b^2/2.  Thus, b <-
2*coef(fit0)/coef(fit0), etc.

hope this helps.  spencer graves
questions for yourself, and when it doesn't, it can make it easier for

Patrick Giraudoux wrote:

>Hi,
>
>I am trying to bootstrap the difference between each parameters among two non linear regression (distributed loss model) as
>following:
>
># data.frame
>
>
>>Raies[1:10,]
>>
>>
>   Tps  SolA  Solb
>1    0 32.97 35.92
>2    0 32.01 31.35
>3    1 21.73 22.03
>4    1 23.73 18.53
>5    2 19.68 18.28
>6    2 18.56 16.79
>7    3 18.79 15.61
>8    3 17.60 13.43
>9    4 14.83 12.76
>10   4 17.33 14.91
>etc...
>
># non linear model (work well)
>
>RaiesLossA.nls<-nls(SolA~a/(1+b*Tps)^c,start=c(a=32,b=0.5,c=0.6))
>RaiesLossB.nls<-nls(Solb~a/(1+b*Tps)^c,start=c(a=33,b=1.5,c=0.5))
>
>
># bootstrap
>library(boot)
>
>dif.param<-function(data,i){
>RaiesLossA.nls<-nls(SolA[i]~a/(1+b*Tps[i])^c,start=c(a=31,b=0.5,c=0.6))
>RaiesLossB.nls<-nls(Solb[i]~a/(1+b*Tps[i])^c,start=c(a=33,b=1.4,c=0.5))
>RaiesLossA.nls\$m\$getPars()-RaiesLossB.nls\$m\$getPars()
>}
>
>
>
>> myboot<-boot(Raies,dif.param,R=1000)
>>
>>
>
>Error in numericDeriv(form[], names(ind), env) :
>        Missing value or an Infinity produced when evaluating the model
>
>It seems that the init values (start=) may come not to be suitable while bootstraping. Data can be sent offline to whom wanted to
>try on the dataset.
>
>Any hint welcome!
>
>Best regards,
>
>Patrick Giraudoux
>
>
>University of Franche-Comté
>Department of Environmental Biology
>EA3184 af. INRA
>F-25030 Besançon Cedex
>
>tel.: +33 381 665 745
>fax.: +33 381 665 797
>http://lbe.univ-fcomte.fr
>
>______________________________________________
>R-help at stat.math.ethz.ch mailing list
>https://www.stat.math.ethz.ch/mailman/listinfo/r-help