[R] nls

[Jost Burkardt]

Xiaodong Jin <close2ceo at yahoo.com> writes:

>> y
> [1]   1  11  42  64 108 173 214
>    
>   > t
> [1] 1 2 3 4 5 6 7
>    
>   > nls(1/y ~ c*exp(-a*b*t)+1/b, start=list(a=0.001,b=250,c=5), trace=TRUE)
> 29.93322 :    0.001 250.000   5.000 
> Error in numericDeriv(form[[3]], names(ind), env) : 
>         Missing value or an infinity produced when evaluating the model
>    
>   # the start value for b is almost close to final estimates, 
>   # a is usually small positive close to 0
>   # c is a positive number
>   # Do I need different estimation method?
>    
>   Regards
>   xj

nls fails because it tests large negative parameter value for b (maybe
because of the product a*b). I got around this by using this
reparametrized version

 nls(1/y ~ c*exp(-a2*t)+b2, start=list(a2=0.001*250,b2=1/250,c=5), trace=TRUE)

 Nonlinear regression model
   model:  1/y ~ c * exp(-a2 * t) + b2 
    data:  parent.frame() 
          a2          b2           c 
  2.49029199  0.01010844 11.94220446 
  residual sum-of-squares: 0.0001228177


Regards,
Jost