[R-sig-Geo] Hosmer-Lemeshow and Le Cessie-van Houwelingen

[Komine]

Hi
I would like your help to understand these results of Hosmer-Lemeshow and Le
Cessie-van Houwelingen test. The first invalidate my model and the second
validate it.

reg<-glm(Ignition~FMC,data=Fire,family=binomial)
summary(reg)

Call:
glm(formula = Ignition ~ FMC, family = binomial, data = Fire)
Deviance Residuals: 
     Min        1Q    Median        3Q       Max  
-2.20116  -0.03811   0.09321   0.31675   2.06383  
Coefficients:
            Estimate Std. Error z value Pr(>|z|)    
(Intercept)  7.29351    1.07182   6.805 1.01e-11 ***
FMC         -0.06973    0.01113  -6.266 3.72e-10 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 313.618  on 230  degrees of freedom
Residual deviance:  94.877  on 229  degrees of freedom
AIC: 98.877

Number of Fisher Scoring iterations: 8

The result of the test is: 

 library(MKmisc)
 HLgof.test(fit = fitted(reg), obs = Ignition,X= cbind(FMC))
$C
        Hosmer-Lemeshow C statistic
data:  fitted(reg) and Ignition 
X-squared = 231, df = 8, p-value < 2.2e-16
$H
        Hosmer-Lemeshow H statistic
data:  fitted(reg) and Ignition 
X-squared = 231, df = 8, p-value < 2.2e-16
$gof
        le Cessie-van Houwelingen-Copas-Hosmer global goodness of fit test
data:  fitted(reg) and Ignition 
z = 1.8747, p-value = 0.06084
Messages d'avis :
1: In Ops.factor(1, obs) : - not meaningful for factors
2: In Ops.factor(1, obs) : - not meaningful for factors

1-	In case these codes are correct, what is the reason of contradiction
between Hosmer-Lemeshow and Le Cessie-van Houwelingen test?  
2-	In this case, what is the conclusion: logistic model is validated or
invalidated? 
Thanks for your help. 


--
View this message in context: http://r-sig-geo.2731867.n2.nabble.com/Hosmer-Lemeshow-and-Le-Cessie-van-Houwelingen-tp6279901p6279901.html
Sent from the R-sig-geo mailing list archive at Nabble.com.