Boa tarde pessoal,

 

        Tenho uma váriavel resposta binomial chamada Infestação (1 presença e 0 ausência de insetos) e dois fatores qualitativos que são vegetação (pastagem e floresta) e altitude (alta e baixa), além de parcelle que são as repetições no campo. Gostaria de fazer contrastes utilizando glm (family=binomial), então primeiro eu fiz:

 

> m.full<-glm(Infestacao~Altitude*Vegetation*Parcelle, data=H.termes, family="binomial")# Modelo completo

Mensagens de aviso perdidas:

In model.matrix.default(mt, mf, contrasts) :

  variable 'Altitude' converted to a factor

 

Já fiquei preocupado com essa mensagem, mais de qualquer maneira Altitude é um fator mesmo, então fiz:

 

> summary(m.full)

 

Call:

glm(formula = Infestacao ~ Altitude * Vegetation * Parcelle,

    family = "binomial", data = H.termes)

 

Deviance Residuals:

    Min       1Q   Median       3Q      Max 

-1.0793  -0.4643  -0.4542  -0.2530   2.6296 

 

Coefficients:

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

(Intercept)                              -3.2454     0.6030  -5.382 7.36e-08 ***

Altitudebaixa                             3.0243     0.6322   4.784 1.72e-06 ***

Vegetationpature                          0.9997     0.9678   1.033  0.30161   

Parcelle                                  0.5115     0.2564   1.995  0.04602 * 

Altitudebaixa:Vegetationpature           -3.0214     1.0505  -2.876  0.00403 **

Altitudebaixa:Parcelle                   -0.5256     0.2669  -1.969  0.04893 * 

Vegetationpature:Parcelle                -0.9048     0.4561  -1.984  0.04729 * 

Altitudebaixa:Vegetationpature:Parcelle   0.9421     0.5024   1.875  0.06073 . 

---

Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

 

(Dispersion parameter for binomial family taken to be 1)

 

    Null deviance: 1578.6  on 1532  degrees of freedom

Residual deviance: 1317.2  on 1525  degrees of freedom

AIC: 1333.2

 

Number of Fisher Scoring iterations: 5

 

Não tem superdispersão então fiz uma anova:

 

> anova(m.full,test="Chi")

Analysis of Deviance Table

 

Model: binomial, link: logit

 

Response: Infestacao

 

Terms added sequentially (first to last)

 

 

                             Df Deviance Resid. Df Resid. Dev P(>|Chi|)   

NULL                                          1532     1578.6             

Altitude                      1   94.613      1531     1484.0   < 2e-16 ***

Vegetation                    1  154.761      1530     1329.2   < 2e-16 ***

Parcelle                      1    0.248      1529     1329.0   0.61866   

Altitude:Vegetation           1    6.511      1528     1322.5   0.01072 * 

Altitude:Parcelle             1    1.195      1527     1321.3   0.27440   

Vegetation:Parcelle           1    0.491      1526     1320.8   0.48328   

Altitude:Vegetation:Parcelle  1    3.577      1525     1317.2   0.05857 . 

---

Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Mensagens de aviso perdidas:

In model.matrix.default(object, data = list(Infestacao = c(0, 0,  :

  variable 'Altitude' converted to a factor

 

Então fui reduzindo o modelo ate restarem as variáveis significativas, sendo:

 

> anova(m3,test="Chi")

Analysis of Deviance Table

 

Model: binomial, link: logit

 

Response: Infestacao

 

Terms added sequentially (first to last)

 

 

                    Df Deviance Resid. Df Resid. Dev P(>|Chi|)   

NULL                                 1532     1578.6             

Altitude             1   94.613      1531     1484.0   < 2e-16 ***

Vegetation           1  154.761      1530     1329.2   < 2e-16 ***

Parcelle             1    0.248      1529     1329.0   0.61866   

Altitude:Vegetation  1    6.511      1528     1322.5   0.01072 * 

---

Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Mensagens de aviso perdidas:

In model.matrix.default(object, data = list(Infestacao = c(0, 0,  :

  variable 'Altitude' converted to a factor

> summary(m3)

 

Call:

glm(formula = Infestacao ~ Altitude + Vegetation + Parcelle +

    Altitude:Vegetation, family = "binomial", data = H.termes)

 

Deviance Residuals:

    Min       1Q   Median       3Q      Max 

-1.0824  -0.4636  -0.4554  -0.3072   2.4807 

 

Coefficients:

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

(Intercept)                    -2.20627    0.23768  -9.283  < 2e-16 ***

Altitudebaixa                   1.91926    0.21846   8.785  < 2e-16 ***

Vegetationpature               -0.83836    0.36768  -2.280  0.02260 * 

Parcelle                        0.01484    0.06585   0.225  0.82173   

Altitudebaixa:Vegetationpature -1.10337    0.41007  -2.691  0.00713 **

---

Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

 

(Dispersion parameter for binomial family taken to be 1)

 

    Null deviance: 1578.6  on 1532  degrees of freedom

Residual deviance: 1322.5  on 1528  degrees of freedom

AIC: 1332.5

 

Number of Fisher Scoring iterations: 5

 

 

  Então posso deduzir, se é que agora fiz tudo certo, que a variável resposta é diferente entre altitude alta e baixa e entre vegetação de pastagem e floresta,  agora gostaria de comparar altitude alta com as diferentes vegetações, só que não entendi porque a análise separou Vegetationbaixa e Vegetationpaturage e também não sei que fator a análise fixou no intercepto,

Alguém saberia me explicar o que posso estar fazendo de errado,

Obrigado,

 

 

 

 

 

Alexandre dos Santos

Ingenieur forestier, Msc.

INRA- Biostatistique et Processus Spatiaux (BioSP)

Domaine Saint-Paul
Site Agroparc
84914 -  Avignon - France
Tél. : +33 (0)6 87 95 16 29