[R] Confidence intervals nls

[Peter Ehlers]

On 2010-02-15 12:01, FishR wrote:
>
> Dear All
>
> I am quite new to R and would appreciate some help fitting 95% confidence
> intervals to a nls function. I have the data
>
> DOY       CET
> 90	5.9
> 91	8
> 92	8.4
> 93	7.7
> 95	6.6
> 96	6.8
> 97	7.1
> 98	9.7
> 99	12.3
> 100	12.8
> 102	11
> 103	9.3
> 104	9.8
> 105	9.9
> 107	7.7
> 110	6.2
> 111	5.9
> 112	5.9
> 113	3.4
> 114	3.5
> 116	3.3
> 117	5.4
> 118	6.3
> 119	9.7
> 120	11.2
> 121	7.3
> 124	7.8
> etc
>
> I am trying to use some code that has been previously posted on the help
> boards but keep getting an error message "dim(X) must have a positive
> length"
>
> plot(DOY, CET)
> model<-nls(CET~a+(b*sin(((2*pi)/365)*(DOY+t))),start=list(a=9.5, b=-6.5,
> t=65))
> summary(model)
> days<-seq(0,365,1)
> predict(model,list(DOY=days))
>
> se.fit<- sqrt(apply(attr(predict(model,list(DOY = days)),"gradient"),1,
>                    function(x) sum(vcov(fm)*outer(x,x))))
> matplot(days, predict(model,list(DOY = days))+
>                 outer(se.fit,qnorm(c(.5, .025,.975))),type="l")
>
> Any help would be greatly appreciated
>
> Best wishes
>
> Tom
>
Maybe you just need to change 'fm' to 'model' in

    function(x) sum(vcov(fm)*outer(x,x))))

?

  -Peter Ehlers