| Author |
Comment/Response |
David Paul
|
 05/10/00 05:47am
I am writing some code to transform bivariate skewed data to approximate multivariate normality using Box-Cox transformations.
It has been no real problem to write functions/procedures which work on bivariate data. However, I have had difficulty figuring out how to extend my functions to handle data of any dimension. Part of this has to do with using the FindMinimum[f,{x,x0},...] command on the LogLikelihood function.
If I know that the data is bivariate in advance, then I simply use FindMinimum[-LogLikelihood,{x,x0},{y,y0}] and obtain the parameter estimates. How do I extend this to data of any dimension? My problem is in figuring out how to specify the {x,x0},{y,y0},... initial starting points, since I don't want to make assumptions about how many dimensions=starting points I need in advance.
Thanks! -Dave
URL: , |
|
