On 10/28/13 00:10, Hansol Yu wrote:> data(Boston, package='MASS')
> y <- Boston$nox
> x <- Boston$dis
> nls(y~ A + B * exp(C * x), start=list(A=1, B=1, C=1))
>
> Error in nls(y ~ A + B * exp(C * x), start = list(A = 1, B = 1, C = 1), :
> step factor 0.000488281 reduced below 'minFactor' of 0.000976562
>
> I don't know how to fix this error. I think my problem is that I set
the
> wrong start. Could somebody help please?
Different starting values will indeed fix the problem. Did you *try*
different starting
values? Take some initiative!
It is pretty clear that C has to be negative, so make the starting value
negative.
Using 1, 1, -1 still doesn't work, but 0.5, 0.5, -0.5 does work.
The function optim() is more robust to starting values. Try
foo <- function(par,x,y){
A <- par[1]
B <- par[2]
C <- par[3]
sum((y - (A+B*exp(C*x)))^2)
}
optim(c(1,1,-1),foo,x=x,y=y,method="BFGS")
The estimates given by optim() are the same as those given by nls() with the
0.5, 0.5, -0.5 starting values.
cheers,
Rolf Turner