sas >> Box-Cox

by renaud.bourles » Wed, 02 Jul 2003 16:41:05 GMT

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

Hi,
I'm currently working on an price index as a trainee with the GREQAM (Marseille, France).
But I don't know how to obtain the likelihood for a linear regression model. In fact, I should use Box-Cox
transformation, but all the macro I've found use the LABR procedure which is not licensed for my system.
So, may somebody help me in teaching me how to program Box-Cox transformation (non autoregressive) or
simply how to find a model likelihood.
Thank you. BOURLES Renaud

Plus simple, plus fiable, plus rapide : douvrez le nouveau Caramail - http://www.caramail.lycos.fr



sas >> Box-Cox

by stringplayer_2 » Thu, 03 Jul 2003 01:19:06 GMT


Renaud,

If you have version 8.2 of SAS or higher, then you can use the
procedure TRANSREG to determine the optimal transformation
employing Box-Cox. The Box-Cox transformation can be obtained
employing the code

proc transreg data=mydata ss2 details;
model boxcox(respnse) = identity(x);
run;

There are parameters which can be specified along with the
boxcox transformation. The following URL links to all of the
SAS version 8.2 documentation pertaining to Box-Cox.

http://support.sas.com/rnd/app/da/new/802ce/stat/chap15/index.htm

Best wishes,

Dale





=====
---------------------------------------
Dale McLerran
Fred Hutchinson Cancer Research Center
mailto: XXXX@XXXXX.COM
Ph: (206) 667-2926
Fax: (206) 667-5977
---------------------------------------

__________________________________
Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!
http://sbc.yahoo.com

sas >> Box-Cox

by Jay Weedon » Thu, 03 Jul 2003 04:02:39 GMT

On 2 Jul 03 17:19:06 GMT, XXXX@XXXXX.COM (Dale McLerran)



Footnote: Box-Cox can AFAIK be applied only to the response variable
in TRANSREG.

JW

Similar Threads

1. Box-Cox (French)

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

Bonjour je suis actuellement en stage au Greqam (Marseille, France) et je construis un indice des prix sur les
logements Marseillais gre au logiciel SAS. Seulement, j'ai besoin pour rliser bien mon travail d'appliquer la
mhode BOX-Cox.
Or, je connais certains probles concernant le calcul de la vraisemblance d'un mode de rression linire.
De plus, j'ai trouver plusieurs macro effectuant Box-Cox mais celles-ci utilisent la procure LABR qui je crois
n'est pas reconnu par ma version (v8). Merci d'avance.
BOURLES Renaud

Pour gagner une Playstation 2, envoyez un SMS avec le code PS au 61321 (0,35 euro hors co du SMS).


2. help with Box-Cox

3. Obtaining box-cox power via univariate procs

Dear All:

I want to obtain the box cox power of a variable I am working with.  I
went via 'proc transreg' but I think that is not the right route
because my intent is to obtain lambda(conv or best) from a univariate
analysis.

Here my  transreg route:

proc transreg data=boxcox ;
 	   model boxcox(x / lambda=-2 to 2 by 0.25) = identity(y);

Does anybody know who I can obtain lambda 'unvariately' for a variable
x rather than modeling against y ?

Thanks,


4. Can someone explain PROC TRANSREG's BoxCox outputs

5. COX-BOX question

  Hi all
  I am wondering anyone know how to use cox-box model by sas software?
Any suggestion will be appreciated.Thanks.

6. AW: Question with time-varying cox model output

7. Question with time-varying cox model output

Hi all,

I am running a cox model with several covariates+a time-varying covariate over a data set, which has missing for categorical variable called var1.

However, "Summary of the Number of Event and Censored Values" from 1)are very different from 2). the first model has much higher numbers than 2nd.

1)
proc tphreg data=dummies covs(aggregate);

           title "Cox model with 3 level dummies";
                  class var1;
                   model time*&censor(0)=var1 var2 breastt/rl ties=efron;
                   breastt=event_breast*time; /***time varying data***/
                   id fam;
                   run;


                       Summary of the Number of Event and Censored Values

                                                               Percent
                             Total       Event    Censored    Censored

                              1422          87        1335       93.88

2)
proc tphreg data=dummies covs(aggregate);

           title "Cox model with 3 level dummies";
                  class var1;
                   model time*&censor(0)=var1 var2/rl ties=efron;
                           id fam;
                   run;

  Summary of the Number of Event and Censored Values

                                                               Percent
                             Total       Event    Censored    Censored

                               689          31         658       95.50

I am not sure why 1st model did not recognize that there are missings in var1?

thanks,

Vincent

8. cox model and multicollinearity