[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-gsl] Multidimensional linear fit (and principal component anal
Barrett C. Foat
Re: [Help-gsl] Multidimensional linear fit (and principal component analysis, covariance matrix)
Sun, 14 Dec 2008 08:03:53 -0600
GSL has multivariate linear regression too:
On Saturday 13 December 2008, Philipp Klaus Krause wrote:
> I want to do a least squares fit of a line in 3 or 4-dimensional space
> to 16 data points.
> I looked at the manual, it seems gsl provides least squares linear fits
> only for onedimensional stuff.
> The classic way would be the principal component analysis (PCA), again
> gsl does not provide this. PCA can be done by estimating the covariance
> matrix and getting thet eigenvector for the biggest eigenvalue. gsl
> seems to provide functions to get the eigenvalues and vectors, it even
> sorts them for me. It might be a bit inefficient to calculate them all
> when I need only the biggest, but that shouldn't be much of a problem.
> However gsl seems to provide estimation of covariance only in one
> dimension, so I would have to implement estimation of the covariance
> matrix myself.
> Is this correct? Will performance be okay for such small data sets (16
> data points, in 3 or 4 dimensions) or is gsl optimized too much towards
> "bigger" problems?
> Help-gsl mailing list