[R] first value from nlm (non-finite value supplied by nlm)

[Spencer Graves]

Have you tried fn.2(out$estimate), without the call to "nlm"?  There may 
be a problem with the definition of "fn.2", but I don't have time to 
check that for you, other than to note the "t" is the matrix transpose 
function.  In many but not all contexts, R is able to distinguish t as a 
function from t as a non-function.  However, I consider that skating on 
thin ice. 

hope this helps.  spencer graves

Thomas Bock wrote:

> Dear expeRts,
>
> first of all I'd like to thank you for the
> quick help on my last which() problem.
>
> Here is another one I could not tackle:
> I have data on an absorption measurement which I want to fit
> with an voigt profile:
>
>  fn.1 <- function(p){
>    for (i1 in ilong){
>      ff <- f[i1]
>      ex[i1] <- exp(S*n*L*voigt(u,v,ff,p[1],p[2],p[3])[[1]])
>    }
> sum((t-ex)^2)
>  }
> out <- nlm(fn.1, p = c(fo, GG, GL), hessian = TRUE, steptol = 1e-5, 
> iterlim = 1000)
> foN <-  out$estimate[1]
> GGN <-  out$estimate[2]
> GLN <-  out$estimate[3]
>
> This works fine but the my start value of S is to poor,
> so I like to fit S in a second run, with the initial values from the 
> first run
> (two runs because I know that S as an parameter is an problem):
>
>  fn.2 <- function(p){
>    for (i1 in ilong){
>      ex[i1] <- exp(p[1]*n*L*voigt(u,v,f[i1],p[2],p[3],GLN)[[1]])
>          }
> sum((t-ex)^2)
> }
> out <-nlm(fn.2, p = c(S,foN,GGN), hessian = TRUE,
>           steptol = 1e-5, iterlim = 1000,print.level=2)
> SN <-  out$estimate[1]
>
> The problem is now that the first value from nlm() is positive (1E-6 !?)
> and this leeds to an Inf:
>
>  iteration = 0
> Step:
> [1] 0 0 0
> Parameter:
> [1] -3.800000e-19  2.196660e+03  5.211179e-03
> Function Value
> [1] 0.5890603
> Gradient:
> [1]       Inf  11.23381 -23.61961 -
>
> Error in nlm(fn.2, p = c(S, foN, GGN), hessian = TRUE, steptol = 1e-5,  :
>    non-finite value supplied by nlm
> In addition: Warning message:
> NA/Inf replaced by maximum positive value
>
> The number of parameters plays no role; same behaviour with p = c(S,GGN)
>
> Can someone give a broad hint
>             Thomas
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help