Can your multipolyfit module could resolve the regressio problem fo a known form functions?

[clockzhong]

hi, Matthew Rocklin,

I traced your question in stackoverflow from here:
https://stackoverflow.com/questions/10988082/multivariate-polynomial-regression-with-numpy
Because I've a very similar task as yours. But I found your module still couldn't match my requirements. Could you please have a look on my problme also?

1. We know the function form already, e.g. r = a*y**2+b * x * y+c * z
   and the x, y, z is indeterminates(variables), the r is the result, and (a, b, c) are the coeffients.
2. We have a lot of samples of (r, x, y, z), how can we estimate the values of (a, b, c). I found your solution and many other python's libraries only support a general formula regression, instead of a specific formula like this. E.g. my example form: r = a*y**2+b * x * y+c * z, we already know the coefficients of x ** 2, z ** 2 are zeros.

Do you have any good suggesions? or your module already support this solution, please give some hints or suggestion.

Thanks in advance.

Clock