Nous allons importer le Dataset nommé louisevie
> louisevie<-[Link](file=[Link](), header=FALSE, se="")
> dim(louisevie)
[1] 200 7
> names(louisevie)
[1] "V1" "V2" "V3" "V4" "V5" "V6" "V7"
2. Nommons les variables
> colnames(louisevie)<-c("sexe", "age", "abo", "sitfam", "soc", "zau", "sal")
> attach(louisevie)
1. Présentation des statistiques descriptives de la base:
a. Calcul de la moyenne, de l’ecart type, du minimum et du maximum pour l’ensemble des
variables continues de la base :
> summary(louisevie)
sexe age abo sitfam soc
Femme:124 Min. :18.00 Min. :0.000 Min. :1.00 Cadre :32
Homme: 76 1st Qu.:35.00 1st Qu.:0.000 1st Qu.:1.00 Employe:88
Median :44.00 Median :1.000 Median :1.00 Ouvrier:80
Mean :42.52 Mean :0.545 Mean :1.63
3rd Qu.:49.00 3rd Qu.:1.000 3rd Qu.:2.00
Max. :59.00 Max. :1.000 Max. :3.00
zau sal
Min. :1.00 Min. : 3678
1st Qu.:2.00 1st Qu.:13789
Median :2.00 Median :17038
Mean :3.03 Mean :18615
3rd Qu.:4.00 3rd Qu.:22523
Max. :7.00 Max. :35972
b. Calcul de la moyenne, l’ecart-type, le minimum et le maximum pour les variables continues
selon le type d’abonnement
> sd(zau)
[1] 1.90453
> sd(sal)
[1] 7082.492
> summary(louisevie [louisevie [,"abo"]=="0","age"])
Min. 1st Qu. Median Mean 3rd Qu. Max.
31.00 34.00 39.00 40.27 46.00 51.00
> sd(louisevie [louisevie [, "abo"]=="0","age"])
[1] 6.752886
> summary(louisevie [louisevie [,"abo"]=="1","age"])
Min. 1st Qu. Median Mean 3rd Qu. Max.
18.00 38.00 46.00 44.39 52.00 59.00
> sd(louisevie [louisevie [,"abo"]=="1","age"])
[1] 9.628294
> summary(louisevie [louisevie [,"abo"]=="0","zau"])
Min. 1st Qu. Median Mean 3rd Qu. Max.
1.000 2.000 2.000 3.055 4.500 7.000
> sd(louisevie [louisevie [,"abo"]=="0","zau"])
[1] 1.928515
> summary(louisevie [louisevie [,"abo"]=="1","zau"])
Min. 1st Qu. Median Mean 3rd Qu. Max.
1.000 2.000 2.000 3.009 4.000 7.000
> sd(louisevie [louisevie [,"abo"]=="1","zau"])
[1] 1.892947
> summary(louisevie [louisevie [,"abo"]=="0","sal"])
Min. 1st Qu. Median Mean 3rd Qu. Max.
3678 12936 16227 17654 21320 35374
> sd(louisevie [louisevie [,"abo"]=="0","sal"])
[1] 7150.625
> summary(louisevie [louisevie [,"abo"]=="1","sal"])
Min. 1st Qu. Median Mean 3rd Qu. Max.
6296 15528 18190 19418 23030 35972
> sd(louisevie [louisevie [,"abo"]=="1","sal"])
[1] 6956.164
c. Nous allons realiser un tri à plat pour l’ensemble des variables qualitatives
> table(sexe)
sexe
Femme Homme
124 76
> length(sexe)
[1] 200
> [Link](table(sexe))*100
sexe
Femme Homme
62 38
> table(abo)
abo
0 1
91 109
> length(abo)
[1] 200
> [Link](table(abo))*100
abo
0 1
45.5 54.5
> table(sitfam)
sitfam
1 2 3
101 72 27
> legnth(sitfam)
Error in legnth(sitfam) : could not find function "legnth"
> length(sitfam)
[1] 200
> [Link](table(sitfam))*100
sitfam
1 2 3
50.5 36.0 13.5
> table(soc)
soc
Cadre Employe Ouvrier
32 88 80
> length(soc)
[1] 200
> [Link](table(soc))*100
soc
Cadre Employe Ouvrier
16 44 40
> table(zau)
zau
1 2 3 4 5 7
31 97 3 21 24 24
> length(zau)
[1] 200
> [Link](table(zau))*100
zau
1 2 3 4 5 7
15.5 48.5 1.5 10.5 12.0 12.0
> addmargins(table(sexe,abo))
abo
sexe 0 1 Sum
Femme 67 57 124
Homme 24 52 76
Sum 91 109 200
> Tab<-table(sexe,abo)
> addmargins([Link](addmargins(Tab,1),1),2)*100
abo
sexe 0 1 Sum
Femme 54.03226 45.96774 100.00000
Homme 31.57895 68.42105 100.00000
Sum 45.50000 54.50000 100.00000
> Tab<-table(sexe,abo)
c. Realisons un test de correlation que nous allons presenter dans un tableau croise entre les
variables d’abonnement et sexe.
> addmargins([Link](addmargins(Tab,2),2),1)*100
abo
sexe 0 1 Sum
Femme 73.62637 52.29358 62.00000
Homme 26.37363 47.70642 38.00000
Sum 100.00000 100.00000 100.00000
> genre<-ifelse(sexe=="Homme",1,0)
> [Link](genre,abo)
Pearson's product-moment correlation
data: genre and abo
t = 3.1561, df = 198, p-value = 0.001848
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
0.08262675 0.34706143
sample estimates:
cor
0.2188588
2. Créons de nouvelles variables
> abon<-ifelse(abo=="0",1,0)
> fem<-ifelse(sexe=="Femme",1,0)
> hom<-ifelse(sexe=="Homme",1,0)
> marie<-ifelse(sitfam=="1",1,0)
> celib<-ifelse(sitfam=="2",1,0)
> divor<-ifelse(sitfam=="3",1,0)
> cadre<-ifelse(soc=="Cadre",1,0)
> empl<-ifelse(soc=="Employe",1,0)
> ouv<-ifelse(soc=="Ouvrier",1,0)
>
> urbain<-ifelse(zau<="3",1,0)
> Femu<-ifelse(sexe=="Femme" & urbain=="1",1,0)
> Femr<-ifelse(sexe=="Femme" & urbain=="0",1,0)
> Homu<-ifelse(sexe=="Homme" & urbain=="1",1,0)
> Homr<-ifelse(sexe=="Homme" & urbain=="0",1,0)
> saldiv<-sal/1000
> saldiv2<-(sal^2)/100000
> logsal<-log(sal)
> age2<-age^2
> logage<-log(age
> table(abo + abon)
3. Vérifions alors que ces variables ont été correctement créées
1
200
> table(fem + hom)
1
200
> table(marie + celib + divor)
1
200
> table(cadre + ouv + empl)
1
200
> cbind(zau, urbain)
zau urbain
[1,] 3 1
[2,] 3 1
[3,] 3 1
[4,] 4 0
[5,] 4 0
[6,] 4 0
[7,] 4 0
[8,] 4 0
[9,] 4 0
[10,] 4 0
[11,] 4 0
[12,] 4 0
[13,] 4 0
[14,] 2 1
[15,] 2 1
[16,] 2 1
[17,] 1 1
[18,] 1 1
[19,] 1 1
[20,] 1 1
[21,] 1 1
[22,] 1 1
[23,] 1 1
[24,] 1 1
[25,] 2 1
[26,] 2 1
[27,] 2 1
[28,] 2 1
[29,] 2 1
[30,] 2 1
[31,] 2 1
[32,] 2 1
[33,] 2 1
[34,] 2 1
[35,] 2 1
[36,] 2 1
[37,] 2 1
[38,] 2 1
[39,] 2 1
[40,] 2 1
[41,] 2 1
[42,] 2 1
[43,] 2 1
[44,] 2 1
[45,] 2 1
[46,] 2 1
[47,] 2 1
[48,] 5 0
[49,] 5 0
[50,] 5 0
[51,] 5 0
[52,] 2 1
[53,] 2 1
[54,] 2 1
[55,] 2 1
[56,] 2 1
[57,] 2 1
[58,] 2 1
[59,] 2 1
[60,] 2 1
[61,] 2 1
[62,] 2 1
[63,] 2 1
[64,] 2 1
[65,] 2 1
[66,] 2 1
[67,] 2 1
[68,] 2 1
[69,] 2 1
[70,] 2 1
[71,] 2 1
[72,] 2 1
[73,] 2 1
[74,] 2 1
[75,] 2 1
[76,] 2 1
[77,] 2 1
[78,] 2 1
[79,] 5 0
[80,] 5 0
[81,] 5 0
[82,] 5 0
[83,] 5 0
[84,] 5 0
[85,] 5 0
[86,] 5 0
[87,] 7 0
[88,] 1 1
[89,] 1 1
[90,] 1 1
[91,] 7 0
[92,] 7 0
[93,] 7 0
[94,] 7 0
[95,] 7 0
[96,] 2 1
[97,] 1 1
[98,] 1 1
[99,] 2 1
[100,] 2 1
[101,] 1 1
[102,] 1 1
[103,] 2 1
[104,] 1 1
[105,] 2 1
[106,] 2 1
[107,] 5 0
[108,] 1 1
[109,] 5 0
[110,] 5 0
[111,] 5 0
[112,] 1 1
[113,] 5 0
[114,] 5 0
[115,] 7 0
[116,] 7 0
[117,] 7 0
[118,] 7 0
[119,] 7 0
[120,] 7 0
[121,] 7 0
[122,] 7 0
[123,] 7 0
[124,] 7 0
[125,] 7 0
[126,] 7 0
[127,] 7 0
[128,] 1 1
[129,] 1 1
[130,] 1 1
[131,] 5 0
[132,] 5 0
[133,] 1 1
[134,] 1 1
[135,] 1 1
[136,] 1 1
[137,] 5 0
[138,] 5 0
[139,] 5 0
[140,] 4 0
[141,] 4 0
[142,] 1 1
[143,] 4 0
[144,] 4 0
[145,] 4 0
[146,] 4 0
[147,] 4 0
[148,] 7 0
[149,] 5 0
[150,] 2 1
[151,] 2 1
[152,] 2 1
[153,] 2 1
[154,] 2 1
[155,] 2 1
[156,] 2 1
[157,] 2 1
[158,] 7 0
[159,] 1 1
[160,] 7 0
[161,] 1 1
[162,] 7 0
[163,] 7 0
[164,] 4 0
[165,] 1 1
[166,] 1 1
[167,] 4 0
[168,] 4 0
[169,] 4 0
[170,] 2 1
[171,] 2 1
[172,] 2 1
[173,] 2 1
[174,] 2 1
[175,] 2 1
[176,] 2 1
[177,] 2 1
[178,] 1 1
[179,] 2 1
[180,] 2 1
[181,] 2 1
[182,] 2 1
[183,] 2 1
[184,] 2 1
[185,] 2 1
[186,] 2 1
[187,] 2 1
[188,] 2 1
[189,] 2 1
[190,] 2 1
[191,] 2 1
[192,] 2 1
[193,] 2 1
[194,] 2 1
[195,] 2 1
[196,] 2 1
[197,] 2 1
[198,] 2 1
[199,] 2 1
[200,] 2 1
> table((Femu + Femr) - fem)
0
200
> table((Homu + Homr) - hom)
0
200
> table(saldiv*1000 - sal)
-9.09494701772928e-13 0 1.81898940354586e-12
1 197 2
> table(sqrt(saldiv2*100000) - sal)
-3.63797880709171e-12 -1.81898940354586e-12 0
1 7 184
9.09494701772928e-13 1.81898940354586e-12 3.63797880709171e-12
1 4 3
> table(sqrt(age2) – age)
Error: unexpected input in "table(sqrt(age2) –"
> table(sqrt(age2)- age)
0
200
> table(exp(logage)- age)
-1.4210854715202e-14 -7.105427357601e-15 -3.5527136788005e-15
4 83 7
0 3.5527136788005e-15 7.105427357601e-15
57 5 32
1.4210854715202e-14 2.1316282072803e-14
8 4
4. Estimons le modèle logit
> LogitAboCst<-glm(abo ~ 1, family = binomial(link="logit"))
> summary(LogitAboCst)
Call:
glm(formula = abo ~ 1, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-1.255 -1.255 1.102 1.102 1.102
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.1805 0.1420 1.271 0.204
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 275.64 on 199 degrees of freedom
AIC: 277.64
Number of Fisher Scoring iterations: 3
> BetaAboCst<-coef(LogitAboCst)
> P1<-exp(BetaAboCst)/(1+exp(BetaAboCst))
> LogitAbonCst<-glm(abon ~ 1, family = binomial(link="logit"))
> summary(LogitAbonCst)
Call:
glm(formula = abon ~ 1, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-1.102 -1.102 -1.102 1.255 1.255
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.1805 0.1420 -1.271 0.204
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 275.64 on 199 degrees of freedom
AIC: 277.64
Number of Fisher Scoring iterations: 3
> BetaAbonCst<-coefficients(LogitAbonCst) P0<-exp(BetaAbonCst)/(1+exp(BetaAbonCst))
Error: unexpected symbol in "BetaAbonCst<-coefficients(LogitAbonCst) P0"
> BetaAbonCst<-coefficients(LogitAbonCst)
> P0<-exp(BetaAbonCst)/(1+exp(BetaAbonCst))
> P0
(Intercept)
0.455
6. Estimation avec variables explicatives
Une seule variable explicative : le sexe
> LogitSexe<-glm(abo ~ fem, family = binomial(link="logit"))
> summary(LogitSexe)
Call:
glm(formula = abo ~ fem, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-1.5183 -1.1096 0.8712 1.2468 1.2468
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.7732 0.2468 3.133 0.00173 **
fem -0.9348 0.3056 -3.059 0.00222 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 265.89 on 198 degrees of freedom
AIC: 269.89
Number of Fisher Scoring iterations: 4
> BetaSexe<-coef(LogitSexe)
> Pfem<-exp(BetaSexe[1] + BetaSexe[2])/(1+exp(BetaSexe[1] + BetaSexe[2]))
> Pfem
(Intercept)
0.4596774
> BetaSexe<-coef(LogitSexe)
> Phom<-exp(BetaSexe[1])/(1+exp(BetaSexe[1]))
> Phom
(Intercept)
0.6842105
> OddH<-Phom/(1-Phom)
> OddH
(Intercept)
2.166667
> OddF<-Pfem/(1-Pfem)
> OddF
(Intercept)
0.8507463
5. Estimation avec variables explicatives
> Logitmodel1<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age + sal, family =
binomial(link="logit")) summary(Logitmodel1)
Error: unexpected symbol in "Logitmodel1<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age +
sal, family = binomial(link="logit")) summary"
> Logitmodel1<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age + sal, family =
binomial(link="logit"))
> summary(Logitmodel1)
Call:
glm(formula = abo ~ fem + marie + celib + cadre + empl + urbain +
age + sal, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.3577 -0.8211 0.3776 0.7868 1.6585
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -2.736e+00 1.145e+00 -2.389 0.01687 *
fem -1.141e+00 3.816e-01 -2.989 0.00280 **
marie 1.209e+00 4.929e-01 2.453 0.01417 *
celib -4.820e-01 5.283e-01 -0.912 0.36156
cadre 1.753e+00 6.443e-01 2.721 0.00652 **
empl -5.122e-01 3.950e-01 -1.297 0.19470
urbain -1.486e-01 3.648e-01 -0.408 0.68363
age 8.629e-02 2.211e-02 3.902 9.55e-05 ***
sal -1.855e-05 2.753e-05 -0.674 0.50044
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 209.36 on 191 degrees of freedom
AIC: 227.36
Number of Fisher Scoring iterations: 4
> Logitmodel2<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age + saldiv, family =
binomial(link="logit"))
> summary(Logitmodel2)
Call:
glm(formula = abo ~ fem + marie + celib + cadre + empl + urbain +
age + saldiv, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.3577 -0.8211 0.3776 0.7868 1.6585
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -2.73595 1.14502 -2.389 0.01687 *
fem -1.14055 0.38161 -2.989 0.00280 **
marie 1.20914 0.49293 2.453 0.01417 *
celib -0.48205 0.52833 -0.912 0.36156
cadre 1.75291 0.64432 2.721 0.00652 **
empl -0.51224 0.39500 -1.297 0.19470
urbain -0.14865 0.36477 -0.408 0.68363
age 0.08629 0.02211 3.902 9.55e-05 ***
saldiv -0.01855 0.02753 -0.674 0.50044
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 209.36 on 191 degrees of freedom
AIC: 227.36
Number of Fisher Scoring iterations: 4
> Logitmodel3<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age + age2 + saldiv + saldiv2,
family = binomial(link="logit"))
> summary(Logitmodel3)
Call:
glm(formula = abo ~ fem + marie + celib + cadre + empl + urbain +
age + age2 + saldiv + saldiv2, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.3084 -0.8019 0.1074 0.8211 1.9342
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 16.4043850 6.4066398 2.561 0.01045 *
fem -1.0203878 0.4067757 -2.508 0.01213 *
marie 0.9201744 0.5174169 1.778 0.07534 .
celib -0.7646153 0.5815115 -1.315 0.18855
cadre 1.5893378 0.6846792 2.321 0.02027 *
empl -0.6905511 0.4203864 -1.643 0.10045
urbain -0.1508788 0.3894397 -0.387 0.69844
age -0.9119657 0.3134578 -2.909 0.00362 **
age2 0.0120903 0.0038443 3.145 0.00166 **
saldiv 0.0878204 0.1356691 0.647 0.51743
saldiv2 -0.0002430 0.0003136 -0.775 0.43827
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 194.72 on 189 degrees of freedom
AIC: 216.72
Number of Fisher Scoring iterations: 6
> Logitmodel4<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + logage + logsal, family =
binomial(link="logit"))
> summary(Logitmodel4)
Call:
glm(formula = abo ~ fem + marie + celib + cadre + empl + urbain +
logage + logsal, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.3864 -0.8067 0.3883 0.8450 1.6142
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -8.9307 5.3756 -1.661 0.096640 .
fem -1.1393 0.3831 -2.974 0.002940 **
marie 1.2387 0.4904 2.526 0.011551 *
celib -0.4824 0.5223 -0.924 0.355692
cadre 1.6911 0.6395 2.644 0.008182 **
empl -0.4704 0.3888 -1.210 0.226252
urbain -0.1317 0.3596 -0.366 0.714158
logage 3.0719 0.8607 3.569 0.000358 ***
logsal -0.2027 0.4904 -0.413 0.679370
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 212.87 on 191 degrees of freedom
AIC: 230.87
Number of Fisher Scoring iterations: 4
> quantile(age, probs=c(0, 0.01 , 0.05, 0.1, 0.25, 0.5 , 0.75 , 0.9, 0.95, 0.99, 1))
0% 1% 5% 10% 25% 50% 75% 90% 95% 99% 100%
18.00 26.95 30.00 31.00 35.00 44.00 49.00 52.20 58.00 59.00 59.00
> age35<-ifelse(age<=35,1,0)
> age45<-ifelse(age>35 & age<=45,1,0)
> age60<-ifelse(age>45,1,0)
> table(age35 + age45 + age60)
1
200
> quantile(sal, probs=c(0, 0.01 , 0.05, 0.1, 0.25, 0.5 , 0.75 , 0.9, 0.95, 0.99, 1))
0% 1% 5% 10% 25% 50% 75% 90% 95% 99% 100%
3678.00 6530.63 8359.00 10725.50 13789.00 17037.50 22522.50 28848.70 33448.90 35787.00
35972.00
> sal15<-ifelse(sal<=15000,1,0)
> sal225<-ifelse(sal>15000 & sal<=22500,1,0)
> sal30<-ifelse(sal>22500 & sal<=30000,1,0)
> salsup<-ifelse(sal>30000,1,0)
> table(sal15 + sal225 + sal30 + salsup)
1
200
> Logitmodel5<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age45 + age60 + sal225 + sal30 +
salsup, family = binomial(link="logit"))
> summary(Logitmodel5)
Call:
glm(formula = abo ~ fem + marie + celib + cadre + empl + urbain +
age45 + age60 + sal225 + sal30 + salsup, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.1555 -0.7690 0.3513 0.8202 1.7381
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.8928 0.7631 -1.170 0.242033
fem -0.9723 0.3896 -2.496 0.012577 *
marie 1.3257 0.5278 2.512 0.012017 *
celib -0.2714 0.5460 -0.497 0.619072
cadre 1.9748 0.6860 2.879 0.003993 **
empl -0.5426 0.3993 -1.359 0.174201
urbain -0.1790 0.3780 -0.474 0.635848
age45 1.0952 0.5124 2.137 0.032582 *
age60 1.7658 0.4564 3.869 0.000109 ***
sal225 0.5910 0.4169 1.418 0.156277
sal30 -0.4469 0.5844 -0.765 0.444442
salsup -0.0087 0.7472 -0.012 0.990711
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 207.24 on 188 degrees of freedom
AIC: 231.24
Number of Fisher Scoring iterations: 4
> Logitmodel6<-glm(abo ~ Femu + Femr + Homu + age + age2 + marie + celib + cadre + empl + saldiv +
saldiv2, family = binomial(link="logit"))
> summary(Logitmodel6)
Call:
glm(formula = abo ~ Femu + Femr + Homu + age + age2 + marie +
celib + cadre + empl + saldiv + saldiv2, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.2200 -0.8103 0.1021 0.8136 1.9393
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 16.2383641 6.4650965 2.512 0.01202 *
Femu -0.9976735 0.5797050 -1.721 0.08525 .
Femr -0.6171168 0.6473748 -0.953 0.34046
Homu 0.2538972 0.6416142 0.396 0.69231
age -0.9216167 0.3167366 -2.910 0.00362 **
age2 0.0122138 0.0038852 3.144 0.00167 **
marie 0.9340910 0.5194527 1.798 0.07214 .
celib -0.8145164 0.5867700 -1.388 0.16510
cadre 1.6176862 0.6859913 2.358 0.01837 *
empl -0.6305470 0.4273594 -1.475 0.14009
saldiv 0.0951634 0.1355617 0.702 0.48268
saldiv2 -0.0002576 0.0003126 -0.824 0.40991
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 194.09 on 188 degrees of freedom
AIC: 218.09
Number of Fisher Scoring iterations: 6
> vcov(Logitmodel6)
(Intercept) Femu Femr Homu age age2 marie celib cadre
empl saldiv
(Intercept) 41.797472688 -3.662931e-01 -4.710581e-01 -0.0045996203 -1.984284e+00 2.409211e-02 -
6.047388e-01 -8.340558e-01 1.585785e-01 -3.438164e-01 -1.220164e-01
Femu -0.366293072 3.360578e-01 2.594732e-01 0.2459708281 4.974367e-03 -9.218309e-05 -
1.370098e-02 -1.656787e-02 -2.904523e-02 -2.360264e-03 4.964772e-03
Femr -0.471058104 2.594732e-01 4.190941e-01 0.2430897581 3.191492e-03 -5.707143e-05 -
8.193036e-03 -3.421170e-02 -4.179343e-02 1.244309e-02 1.836028e-02
Homu -0.004599620 2.459708e-01 2.430898e-01 0.4116688275 -1.122594e-02 1.335229e-04
1.346819e-02 -3.645484e-02 5.003942e-02 4.464801e-02 -5.671508e-03
age -1.984283628 4.974367e-03 3.191492e-03 -0.0112259396 1.003221e-01 -1.226524e-03
1.824812e-02 3.355463e-02 -5.162942e-03 1.625846e-02 -2.901064e-03
age2 0.024092113 -9.218309e-05 -5.707143e-05 0.0001335229 -1.226524e-03 1.509508e-05 -
2.063893e-04 -3.931868e-04 9.527985e-05 -2.190399e-04 3.684612e-05
marie -0.604738801 -1.370098e-02 -8.193036e-03 0.0134681942 1.824812e-02 -2.063893e-04
2.698311e-01 2.032042e-01 -3.418169e-03 2.120221e-02 6.145067e-05
celib -0.834055801 -1.656787e-02 -3.421170e-02 -0.0364548447 3.355463e-02 -3.931868e-04
2.032042e-01 3.442991e-01 -1.367901e-01 -4.855065e-02 -2.791954e-03
cadre 0.158578546 -2.904523e-02 -4.179343e-02 0.0500394168 -5.162942e-03 9.527985e-05 -
3.418169e-03 -1.367901e-01 4.705840e-01 1.217114e-01 -1.132606e-02
empl -0.343816369 -2.360264e-03 1.244309e-02 0.0446480091 1.625846e-02 -2.190399e-04
2.120221e-02 -4.855065e-02 1.217114e-01 1.826361e-01 -4.557175e-03
saldiv -0.122016351 4.964772e-03 1.836028e-02 -0.0056715076 -2.901064e-03 3.684612e-05
6.145067e-05 -2.791954e-03 -1.132606e-02 -4.557175e-03 1.837698e-02
saldiv2 0.000261332 -2.124185e-06 -3.547225e-05 0.0000176759 6.438946e-06 -8.515026e-08
1.506216e-06 8.595541e-06 9.845642e-06 1.028776e-05 -4.147838e-05
saldiv2
(Intercept) 2.613320e-04
Femu -2.124185e-06
Femr -3.547225e-05
Homu 1.767590e-05
age 6.438946e-06
age2 -8.515026e-08
marie 1.506216e-06
celib 8.595541e-06
cadre 9.845642e-06
empl 1.028776e-05
saldiv -4.147838e-05
saldiv2 9.773574e-08
> varFemu<-vcov(Logitmodel6)[2,2]
> varFemr<-vcov(Logitmodel6)[3,3]
> covFemuFemr<-vcov(Logitmodel6)[2,3]
> varFemuFemr<- varFemu + varFemr - 2*covFemuFemr
> seFemuFemr<-sqrt(varFemuFemr)
> BetaFemu<-coef(Logitmodel6)[2]
> BetaFemr<-coef(Logitmodel6)[3]
> Z <- (BetaFemu - BetaFemr)/seFemuFemr
>Z
Femu
-0.7830228
> pvalue1<-2*pt(Z, 200-12, [Link]=TRUE)
> pvaluel
Error: object 'pvaluel' not found
> pvalue1
Femu
0.434599
> Logitmodel7<-glm(abo ~ fem + marie + celib + cadre + empl + urbain + age + age2 + saldiv + saldiv2,
family = binomial(link="logit"))
> summary(Logitmodel7)
Call:
glm(formula = abo ~ fem + marie + celib + cadre + empl + urbain +
age + age2 + saldiv + saldiv2, family = binomial(link = "logit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.3084 -0.8019 0.1074 0.8211 1.9342
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 16.4043850 6.4066398 2.561 0.01045 *
fem -1.0203878 0.4067757 -2.508 0.01213 *
marie 0.9201744 0.5174169 1.778 0.07534 .
celib -0.7646153 0.5815115 -1.315 0.18855
cadre 1.5893378 0.6846792 2.321 0.02027 *
empl -0.6905511 0.4203864 -1.643 0.10045
urbain -0.1508788 0.3894397 -0.387 0.69844
age -0.9119657 0.3134578 -2.909 0.00362 **
age2 0.0120903 0.0038443 3.145 0.00166 **
saldiv 0.0878204 0.1356691 0.647 0.51743
saldiv2 -0.0002430 0.0003136 -0.775 0.43827
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 194.72 on 189 degrees of freedom
AIC: 216.72
Number of Fisher Scoring iterations: 6
> Xmoyen<-cbind(1,0,0,0,0,0,1,42.5, 42.5^2 , 18590/1000 , (18590^2)/100000 )
> Beta7<-coef(Logitmodel7)
> Pmoyen<-exp(Xmoyen%*%Beta7)/(1+exp(Xmoyen%*%Beta7))
> Pmoyen
[,1]
[1,] 0.531402
> Xbis<-cbind(1,0,1,0,1,0,0,50, 50^2 , 30000/1000 , (30000^2)/100000)
> Pbis<-exp(Xbis%*%Beta7)/(1+exp(Xbis%*%Beta7))
> Pbis
[,1]
[1,] 0.9818123
7. Estimer un modèle PROBIT
> ProbitSexe<-glm(abo ~ hom, family = binomial(link="probit"))
> summary(ProbitSexe)
Call:
glm(formula = abo ~ hom, family = binomial(link = "probit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-1.5183 -1.1096 0.8712 1.2468 1.2468
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.1012 0.1128 -0.898 0.36925
hom 0.5808 0.1876 3.096 0.00196 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 265.89 on 198 degrees of freedom
AIC: 269.89
Number of Fisher Scoring iterations: 4
> Prob1<-pnorm(-0.1012 + 0.5808)
> Prob1
[1] 0.6842441
> Prob0<-pnorm(-0.1012)
> Prob0
[1] 0.4596958
> Probitmodel<-glm(abo ~ fem + age + age2 + marie + celib + cadre + empl + urbain + saldiv + saldiv2,
family = binomial(link="probit"))
> summary(Probitmodel)
Call:
glm(formula = abo ~ fem + age + age2 + marie + celib + cadre +
empl + urbain + saldiv + saldiv2, family = binomial(link = "probit"))
Deviance Residuals:
Min 1Q Median 3Q Max
-2.25886 -0.80121 0.05407 0.83502 1.92147
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 9.4792485 3.7221752 2.547 0.01087 *
fem -0.6072863 0.2370584 -2.562 0.01041 *
age -0.5334554 0.1817832 -2.935 0.00334 **
age2 0.0071238 0.0022246 3.202 0.00136 **
marie 0.5736498 0.3099084 1.851 0.06417 .
celib -0.3978892 0.3419977 -1.163 0.24466
cadre 0.8440002 0.3923651 2.151 0.03147 *
empl -0.4432831 0.2491456 -1.779 0.07520 .
urbain -0.0798572 0.2292545 -0.348 0.72759
saldiv 0.0531477 0.0794605 0.669 0.50359
saldiv2 -0.0001479 0.0001833 -0.807 0.41964
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 275.64 on 199 degrees of freedom
Residual deviance: 194.24 on 189 degrees of freedom
AIC: 216.24
Number of Fisher Scoring iterations: 7
