Chapitre 12. R´ egression lin´ eaire simple et multiple
501
> plot(residuals(modele1)∼fitted(modele1),
+
xlab="Valeurs prédites",ylab="Résidus")
●
●
●● ●
●
●
●
●
●
●
●
●
●
● ●
●
● ● ● ●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●●
●
●
● ●
● ●
●
●
● ● ●
●
●
● ●
●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
● ●
●
● ●
●
●
●
● ● ●
●
●
●
●
●
●
● ●
●
●
● ●
●
●
●
● ●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
● ● ●
●
●
●●
●
● ●
●
● ● ●
●
●
● ● ●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
2800
3000
3200
3400
−2000 −1000
0
1000
2000
Valeurs prédites
Résidus
Figure 12.5 – Graphe des r´ esidus en fonction des valeurs pr´ edites.
Remarque
Signalons que vous pouvez obtenir certains graphiques diagnostiques du
mod` ele en tapant les instructions suivantes :
> par(mfrow=c(1,2))
> plot(modele1,1:2,col.smooth="red")
2800
3200
−2000
0 1000 2000
Fitted values
Residuals
●
●
● ●
● ●
●
● ●
●
●
●
●● ● ●
● ● ● ● ●
●
●
● ●
●
●
● ●
● ●
●
●
●
● ●
●
●
●
●
● ●
●
●
● ●
● ●
●
●
● ● ●
● ●
● ●
●
● ● ● ●
●
●
●
●
●
●
● ●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
● ●
●
● ● ●
●
●
● ● ●
● ●
●
●
●
● ●●
●
●
●
●
● ●
●●
●
●
● ●
●
●
●
● ●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
● ● ●
●
●
●
●
● ●
● ●● ● ●
●
●● ●
●
●
● ●
●
●●
●
● ● ●
●
●
●● ● ●
●
●
●●
●
●
●
●
● ●
●
●
●
Residuals vs Fitted
131
130
133
●
●
● ● ●
● ● ●
● ●
●
●
●
● ● ● ●
● ● ● ●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
● ● ●
●
●
● ●
●
●
● ●
● ● ●
●
●
● ●
●
● ● ●
●
●
● ● ● ●
● ●
●
●
●
●
●
●
●
● ●
●
●
● ● ●
●
●
●
●
● ● ● ●
●
●
●
● ●
●
●
● ●
● ● ●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
● ●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ● ● ●
●
● ●
●
●
●
●
● ● ●
●
●
● ●
●
●
● ● ● ●
●
●
●
●
● ● ● ●
● ●
−3
−1
1
3
−3 −2 −1 0 1 2 3
Theoretical Quantiles
Standardized residuals
Normal Q−Q
131
133
130
Notons que l’instruction plot(modele1) peut tracer six graphiques,
dont certains traitent de la d´ etection de valeurs atypiques. Nous y reviendrons en d´ etail dans la section « La r´ egression lin´ eaire multiple ».
501
> plot(residuals(modele1)∼fitted(modele1),
+
xlab="Valeurs prédites",ylab="Résidus")
●
●
●● ●
●
●
●
●
●
●
●
●
●
● ●
●
● ● ● ●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●●
●
●
● ●
● ●
●
●
● ● ●
●
●
● ●
●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
● ●
●
● ●
●
●
●
● ● ●
●
●
●
●
●
●
● ●
●
●
● ●
●
●
●
● ●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
● ● ●
●
●
●●
●
● ●
●
● ● ●
●
●
● ● ●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
2800
3000
3200
3400
−2000 −1000
0
1000
2000
Valeurs prédites
Résidus
Figure 12.5 – Graphe des r´ esidus en fonction des valeurs pr´ edites.
Remarque
Signalons que vous pouvez obtenir certains graphiques diagnostiques du
mod` ele en tapant les instructions suivantes :
> par(mfrow=c(1,2))
> plot(modele1,1:2,col.smooth="red")
2800
3200
−2000
0 1000 2000
Fitted values
Residuals
●
●
● ●
● ●
●
● ●
●
●
●
●● ● ●
● ● ● ● ●
●
●
● ●
●
●
● ●
● ●
●
●
●
● ●
●
●
●
●
● ●
●
●
● ●
● ●
●
●
● ● ●
● ●
● ●
●
● ● ● ●
●
●
●
●
●
●
● ●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
● ●
●
● ● ●
●
●
● ● ●
● ●
●
●
●
● ●●
●
●
●
●
● ●
●●
●
●
● ●
●
●
●
● ●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
● ● ●
●
●
●
●
● ●
● ●● ● ●
●
●● ●
●
●
● ●
●
●●
●
● ● ●
●
●
●● ● ●
●
●
●●
●
●
●
●
● ●
●
●
●
Residuals vs Fitted
131
130
133
●
●
● ● ●
● ● ●
● ●
●
●
●
● ● ● ●
● ● ● ●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
● ● ●
●
●
● ●
●
●
● ●
● ● ●
●
●
● ●
●
● ● ●
●
●
● ● ● ●
● ●
●
●
●
●
●
●
●
● ●
●
●
● ● ●
●
●
●
●
● ● ● ●
●
●
●
● ●
●
●
● ●
● ● ●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
● ●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ● ● ●
●
● ●
●
●
●
●
● ● ●
●
●
● ●
●
●
● ● ● ●
●
●
●
●
● ● ● ●
● ●
−3
−1
1
3
−3 −2 −1 0 1 2 3
Theoretical Quantiles
Standardized residuals
Normal Q−Q
131
133
130
Notons que l’instruction plot(modele1) peut tracer six graphiques,
dont certains traitent de la d´ etection de valeurs atypiques. Nous y reviendrons en d´ etail dans la section « La r´ egression lin´ eaire multiple ».