From: Spencer Graves <spencer.graves_at_pdf.com>

Date: Tue 09 Aug 2005 - 03:52:51 EST

>>severine.erhel@free.fr a écrit :

*>>
*

*>>>hello,
*

*>>>
*

*>>>I work since a few time on R and i wanted to know how to obtain the Wald
*

*>>
*

*>>chi
*

*>>
*

*>>>square value when you make a binary logistic regression. In fact, i have
*

*>>
*

*>>the z
*

*>>
*

*>>>value and the signification but is there a script to see what is the value
*

*>>
*

*>>of
*

*>>
*

*>>>Wald chi square. You can see my model below,
*

*>>>Best regards,
*

*>>>
*

*>>>Séverine Erhel
*

*>>>
*

*>>
*

*>>If you want a global test for several coeff associated with the same
*

*>>variable (e.g., form or criter2 in your example), you can fit the model
*

*>>without the variable and compare the 2 models with a likelihood ratio
*

*>>test (function anova): it is safer than the Wald test.
*

*>>
*

*>>If you really want the Wald test, it is available in different packages:
*

*>>see for example the function wald.test in package aod.
*

*>>
*

*>>Best,
*

*>>
*

*>>Renaud
*

*>>
*

*>>
*

*>>
*

*>>>
*

*>>>
*

*>>>
*

*>>>[Previously saved workspace restored]
*

*>>>
*

*>>>
*

*>>>
*

*>>>
*

*>>>>m3 = glm(reponse2 ~ form + factor(critere2)
*

*>>
*

*>>,family=binomial,data=mes.donnees)
*

*>>
*

*>>>>summary (m3)
*

*>>>
*

*>>>
*

*>>>Call:
*

*>>>glm(formula = reponse2 ~ form + factor(critere2), family = binomial,
*

*>>> data = mes.donnees)
*

*>>>
*

*>>>Deviance Residuals:
*

*>>> Min 1Q Median 3Q Max
*

*>>>-2.5402 0.2064 0.3354 0.4833 1.4177
*

*>>>
*

*>>>Coefficients:
*

*>>> Estimate Std. Error z value Pr(>|z|)
*

*>>>(Intercept) 0.5482 0.3930 1.395 0.1631
*

*>>>form Illustration 3.2904 0.6478 5.080 3.78e-07 ***
*

*>>>form Texte+illustration 2.6375 0.4746 5.557 2.74e-08 ***
*

*>>>factor(critere2)2 -1.0973 0.5103 -2.150 0.0315 *
*

*>>>factor(critere2)3 -0.9891 0.5107 -1.937 0.0528 .
*

*>>>---
*

*>>>Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1
*

*>>>
*

*>>>(Dispersion parameter for binomial family taken to be 1)
*

*>>>
*

*>>> Null deviance: 227.76 on 218 degrees of freedom
*

*>>>Residual deviance: 162.11 on 214 degrees of freedom
*

*>>>AIC: 172.11
*

*>>>
*

*>>>Number of Fisher Scoring iterations: 5
*

*>>>
*

*>>>______________________________________________
*

*>>>R-help@stat.math.ethz.ch mailing list
*

*>>>https://stat.ethz.ch/mailman/listinfo/r-help
*

*>>>PLEASE do read the posting guide!
*

*>>
*

*>>http://www.R-project.org/posting-guide.html
*

*>>
*

*>>
*

*>>--
*

*>>Dr Renaud Lancelot, vétérinaire
*

*>>Projet FSP régional épidémiologie vétérinaire
*

*>>C/0 Ambassade de France - SCAC
*

*>>BP 834 Antananarivo 101 - Madagascar
*

*>>
*

*>>e-mail: renaud.lancelot@cirad.fr
*

*>>tel.: +261 32 40 165 53 (cell)
*

*>> +261 20 22 665 36 ext. 225 (work)
*

*>> +261 20 22 494 37 (home)
*

*>>
*

Date: Tue 09 Aug 2005 - 03:52:51 EST

Is this a research question? If not, I'd like to know why you think the Wald test is better.

Are you famililiar with Bates and Watts (1988) Nonlinear Regression Analysis and Its Applications (Wiley), and with the concepts of "intrinsic" and "parameter effects" nonlinearity? In brief, nonlinear regression and maximum likelihood estimation more generally involve projection onto a nonlinear manifold, which is subject to intrinsic nonlinearity as well as parameter effects nonlinearity. The Wald test suffers from both types of nonlinearity, while the 2*log(likelihood ratio) procedure suffers from only the intrinsic nonlinearity. Moreover, one of the later chapters in Bates and Watts include a comparison intrinsic and parameter effects nonlinearity in several published nonlinear regression examples. I don't remember the details now, but in all but a few cases, the parameter effects were at least an order of magnitude greater than the intrinsic nonlinearity.

If you are not familiar with Bates and Watts, I highly recommend it. If you are, I could see comparing Wald and 2*log(likelihood ratio) to decide if I want to use Wald in certain applications where 2*log(likelihood ratio) may not be feasible.

If you have evidence raising questions about the above, I'd like to know.

spencer graves

severine.erhel@free.fr wrote:

> th,ks for your help, > > i don't have this package on my R, do you know an other package that have this > test...thanks > > > > > Selon Renaud Lancelot <renaud.lancelot@cirad.fr>: > >

>>severine.erhel@free.fr a écrit :

> > > ______________________________________________ > R-help@stat.math.ethz.ch mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html

-- Spencer Graves, PhD Senior Development Engineer PDF Solutions, Inc. 333 West San Carlos Street Suite 700 San Jose, CA 95110, USA spencer.graves@pdf.com www.pdf.com <http://www.pdf.com> Tel: 408-938-4420 Fax: 408-280-7915 ______________________________________________ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.htmlReceived on Tue Aug 09 03:57:07 2005

*
This archive was generated by hypermail 2.1.8
: Sun 23 Oct 2005 - 15:09:55 EST
*