# [R] Regarding R

Douglas Bates bates at stat.wisc.edu
Thu Jan 30 17:20:04 CET 2003

```Komanduru Sai C <sck2348 at cacs.louisiana.edu> writes:

>  I am a grad student at the university of Louisiana at Lafayette. I
>  have a question. I have an equation with 3 unknowns.  f= P/ (r+p)
>  pow(B). I have values of f for different r's. Can i use R to find
>  the P,p,B values which are constants. The equation is Manderbolts
>  equation.

You may want to use nonlinear least squares to fit the parameters in
this model.  To do so you will need starting values for the parameters
p and B.  Suppose that the starting value for p is 0.1 and for B is 2
and that your data are in a data frame called KSC.

fm =  nls(y ~ 1/((r+p)^B), data = KSC, start = c(p=0.1, B=2),
alg='plinear', trace = TRUE)
coef(fm)

Use of nls assumes that the observations consist of 'signal + noise' and
the 'noise' part has constant variance.  See, for example, Bates and
Watts (1988), "Nonlinear Regression Analysis and Its Applications",
Wiley.

```