ellipse

[Daniel J Sebald]

Søren Hauberg wrote:
> lør, 31 01 2009 kl. 16:58 -0600, skrev Daniel J Sebald:
> 
>>I'd say that a routine like this, as nice as it is and as convenient
>>as it is, shouldn't be in Octave itself based on its lack of
>>generality.  It's very directed toward statistics.  Someone wanting an
>>ellipse for some other reason might have a hard time following the
>>relationship between eigenvectors and eigenvalues and covariance
>>matrices, etc.  "shift" and "level" would leave people wondering.
> 
> 
> Well, there is certainly some truth in these statements.
> 
> Perhaps such a function like this should simply accept a center, and a
> major and a minor axis for the ellipse? From a statistics point of view,
> such an API would still be fairly easy to use.

That is my thinking too; the statistics point of view could be transformed to a general equation fairly easily.  The rotation matrix (i.e., generalization of 1/a^2 and 1/b^2) will always factor in somehow, even if it is internal to the routine.

Having multiple input formats, one of them being the current statistical framework, is fine too.

Dan