Search code examples
rerror-handlingnon-linear-regression

R: power function fitting - error fix and more

I would like to fit a power model of type:

y = b*x(^z)

My data is Animals data from mass packages.

My code is here:

library(MASS)
nls(brain~b*body^z,start = list(b = 0, z = 1),data=Animals)

It has an error I don't know how to fix 

Error in nlsModel(formula, mf, start, wts) : 
  singular gradient matrix at initial parameter estimates

And I want to ask is this right to fit this model use my function.

Thanks for your help!

Solution

  • You have two mistakes. One is the singular gradient due to b=0, the other is the fact that there are two different types of animals that follow completely different distributions.

    Look at the distribution:

    enter image description here Fixing the singular gradient still produces an error:

    > nls(brain~b*body^z,start = list(b = 0.1, z = 1),data=Animals)
Error in numericDeriv(form[[3L]], names(ind), env) : 
  Missing value or an infinity produced when evaluating the model

    Therefore, you should also do this: 

    subset(Animals, ! body > 9000) -> mammals
nls(brain~b*body^z,start = list(b = 0.1, z = 1),data=mammals)
Nonlinear regression model
  model: brain ~ b * body^z
   data: mammals
      b       z 
15.5540  0.6795 
 residual sum-of-squares: 4301588

Number of iterations to convergence: 13 
Achieved convergence tolerance: 3.321e-06