Re: PSPP-BUG: Please report to address@hidden

[Jason Stover]

On Fri, Apr 18, 2008 at 06:51:39AM -0400, Mike Marchywka wrote:
> fwiw, if I understand this, it looks like the simple non-parametric stuff is 
> "unimplmeneted?": ChangeLog: * examine.q Added some of the parametric 
> calculationscommand.def:UNIMPL ("COXREG", ERRO, ERRO, ERRO, ERRO, "Cox 
> proportional hazards regression")command.def:UNIMPL ("NONPAR CORR", ERRO, 
> ERRO, ERRO, ERRO, "Nonparametric correlation")command.def:UNIMPL ("NPAR 
> TESTS", ERRO, ERRO, ERRO, ERRO, "Nonparametric tests")command.def:UNIMPL 
> ("SURVIVAL", ERRO, ERRO, ERRO, ERRO, "Survival analysis") 

Yes, these are unimplemented. 

> If you have links for right censored non-parametric tests I am trying to 
> implement my own code
> for some blinded group separation ( mingling outcome data from various 
> treatment groups and
> trying to separate based on smoothness within group using wavelets). I need 
> some reference
> and evaluation metrics as well as more sensical parametric stuff that I can 
> code myself ( or
> pickup normal/hypergeometric etc distributions from canned code elsewhere).

Usually censoring is important in Survival Analysis, so you might check out any 
book
on that. This reference mentions estimating the hazard function via wavelets:

   http://citeseer.ist.psu.edu/482.html

But when you say you want to 'separate based on smoothness within group', it 
sounds
like you are dealing with a classification problem. If so, there are many 
classifiers
that might work. You could start here:

http://en.wikipedia.org/wiki/Statistical_classification

>...
> Warning: There are multiple cygwin1.dlls on your path

I don't know exactly what's wrong, but I would start by making
sure you have only one cygwin1.dll, and make sure it has all the 
necessary packages in it.

-Jason