Re: polyfit weirdness

[Paul Kienzle]

Miquel Cabanas wrote:

> hi,
>
> On Tue, Feb 11, 2003 at 02:08:37AM -0600, Scott Lamb wrote:
>  
>
> > So first I'm calling polyfit to get the trendline:
> >
> >    l1 = 1e-9 * [ 578.45 545.88 435.87 404.71 365.26 ]
> >    v1 =        [   0.66   0.79   1.40   1.58   1.92 ]
> >    f1 = 299792458 ./ l1
> >    [p1,cv1] = polyfit(f1,v1,1)
> >
> > ...and it doesn't return what I'd expect at all:
> >    
>
> I think we can narrow the problem to be related to the
> left-division in polyfit, could you try running (this is
> what polyfit does),
>
>   n = 1;
>   l = 5;
>
>   x = [ 5.1827e+14  5.4919e+14  6.8780e+14  7.4076e+14 ...
>   8.2076e+14 ]';
>
>   y = [ 0.66000  0.79000  1.40000  1.58000  1.92000 ]';
>
>   X = (x * ones (1, n+1)) .^ (ones (l, 1) * (0 : n));
>   p = X \ y;
>   yf = X * p;
>
> In my working Octave I get
>
>   p =
>
>      -1.4948e+00
>       4.1680e-15
>       
> and (estimated y-values)
>
>   yf =
>
>     0.66529
>     0.79416
>     1.37188
>     1.59262
>     1.92605
>
>  
Could this be a property of newer versions of LAPACK?



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------