non-linear fitting , syntax-problem

[Stefan Neumann]

Hi,


here is a working script using function leasqr() to fit a function to a data-set:

filename = "tmp/non-lin.02.dat";
fid = fopen (filename, "r");   
nc = 3 ;    %Number of columns
nr = 98 ;    %Number of data points
[val,count] = fscanf(fid,'%f %f %f',[nc,nr]);      %Read matrix : x,y,z
val = val';                                        %Transpose
fclose (fid);
x = val(1:nr,1:(nc-1));                            %Independent variables
y = val(1:nr,nc);                                  %Dependent variable
pin=ones(1,3);                                    %There are 7 parameters
F = inline("   p(1)*x(:,1)  +  p(2)*x(:,2) + p(3)   ", "x", "p");
[f,p,kvg,iter,corp,covp,covr,stdresid,Z,r2]=leasqr(x,y,pin,F);



and here is one that doesn't work:

filename = "tmp/non-lin.02.dat";
fid = fopen (filename, "r");   
nc = 3 ;    %Number of columns
nr = 98 ;    %Number of data points
[val,count] = fscanf(fid,'%f %f %f',[nc,nr]);      %Read matrix : x,y,z
val = val';                                        %Transpose
fclose (fid);
x = val(1:nr,1:(nc-1));                            %Independent variables
y = val(1:nr,nc);                                  %Dependent variable
pin=ones(1,2);                                    %There are 7 parameters
% Function to fit
F = inline("   p(1)*x(:,1)*x(:,2)+p(2)   ", "x", "p");
[f,p,kvg,iter,corp,covp,covr,stdresid,Z,r2]=leasqr(x,y,pin,F);


The datafile 'tmp/non-lin.02.dat" contains 98 rows of 3 colums each, with column 3 (y) calculated from columns 1+2 (x1 and x2).
(FYI: The script is adapted from the example here: http://knol.google.com/k/multidimensional-non-linear-least-squares-in-octave# )

The first example assumes a linear function y = p1*x1 + p2*x2 + p3

The second example assumes a non-linear function y = p1*x1*x2 + p2

Obviously the second example does not work because of the multiplication of 2 vectors x(:,1) * x(:,2)

What is the correct syntax in this case, or how should I rewrite the script?

TIA, stn