[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## Re: New function

**From**: |
Jason Stover |

**Subject**: |
Re: New function |

**Date**: |
Fri, 27 Mar 2009 17:04:51 -0400 |

**User-agent**: |
Mutt/1.5.18 (2008-05-17) |

On Fri, Mar 27, 2009 at 12:08:25PM -0300, Michel Boaventura wrote:
>* Sorry but I don't understand. Supose that I have a variable V1 who is*
>* categorical and*
>* have three values, 'a','b' and 'c', and another variable V2 who isn't*
>* categorical.*
>* If I have a matrix M like:*
>* *
>* M =*
>* 1 'a'*
>* 2 'c'*
>* 3 'b'*
>* 2 'b'*
>* 5 'c'*
>* 4 'a'*
>* 1 'c'*
>* *
>* How did I use binary vectors to make svd(M)? Because if I choose random values*
>* to represents a,b and c, for each value I will get a different svd.*
You will get a different SVD and covariance matrix. But you should
choose methods of inference that won't depend on this arbitrary frame
of reference. For example, any (reasonable) scheme that sends the
values of V2 to the set of 2-D binary vectors will give the same sums
of squares in analysis of variance.