Exemples-Sélection de modèles 04.11.

19 07:45

> # Librairies utilisées

> # --------------------
>
> library(lmtest) # waldtest, coeftest
Le chargement a nécessité le package : zoo

Attachement du package : ‘zoo’

The following objects are masked from ‘package:base’:

as.Date, as.Date.numeric

> library(memisc) # mtable

Le chargement a nécessité le package : lattice
Le chargement a nécessité le package : MASS

Attachement du package : ‘memisc’

The following objects are masked from ‘package:stats’:

contr.sum, contr.treatment, contrasts

The following object is masked from ‘package:base’:

as.array

>
>
>
> # 1. Données de base
> # ------------------
>
> # a) Lecture des données de base
>
> Bwages <- read.csv2(file(paste(ddpath,"Bwages.csv",sep=""),encoding="latin1"),
+ header=T,sep=";", dec=".")
> head(Bwages)
wage educ exper sex
1 313.8528 1 23 1
2 194.3780 1 15 0
3 426.1364 1 31 1
4 284.0909 1 32 1
5 318.1818 1 9 1
6 330.7895 1 15 0
>
> # b) Facteurs
>
> Bwages <- transform(Bwages,
+ educ = factor(educ, labels=c("1-bas", "2", "3", "4", "5-élevé")),
+ sex = factor(sex, labels=c("Femme", "Homme"))
+ )
>
> # 2. Analyse de régression
> # ------------------------
>
> # a) Formes structurelles
>
> f0 <- formula(log(wage) ~ 1.)
> f1 <- formula(log(wage) ~ educ + exper + I(exper^2))
> f2 <- update(f1, .~. + sex)
> f3 <- update(f1, .~. + sex*educ)
> f4 <- update(f1, .~. + sex*educ + sex*exper)
>
> # a) Estimation des modèles

Page 1 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

>
> # Modèle nul
>
> lm.fit0 <- lm(f0, data = Bwages, na.action = na.exclude)
> summary(lm.fit0)

Call:
lm(formula = f0, data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.55005 -0.24094 -0.01922 0.21156 1.52792

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6.031735 0.009449 638.3 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.3625 on 1471 degrees of freedom

>
> # Modèle 1
>
> lm.fit1 <- lm(f1, data = Bwages, na.action = na.exclude)
> summary(lm.fit1)

Call:
lm(formula = f1, data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.60391 -0.15586 -0.00126 0.17378 1.07186

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.262e+00 3.809e-02 138.173 < 2e-16 ***
educ2 1.412e-01 3.408e-02 4.142 3.63e-05 ***
educ3 2.967e-01 3.283e-02 9.039 < 2e-16 ***
educ4 4.529e-01 3.372e-02 13.433 < 2e-16 ***
educ5-élevé 6.290e-01 3.401e-02 18.496 < 2e-16 ***
exper 3.530e-02 2.583e-03 13.666 < 2e-16 ***
I(exper^2) -4.999e-04 6.515e-05 -7.672 3.06e-14 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.2868 on 1465 degrees of freedom

Multiple R-squared: 0.3767, Adjusted R-squared: 0.3741
F-statistic: 147.6 on 6 and 1465 DF, p-value: < 2.2e-16

>
> # Modèle 2
>
> lm.fit2 <- lm(f2, data = Bwages, na.action = na.exclude)
> summary(lm.fit2)

Call:
lm(formula = f2, data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.51821 -0.15173 0.01222 0.16646 1.09148

Coefficients:

Page 2 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

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

(Intercept) 5.193e+00 3.855e-02 134.708 < 2e-16 ***
educ2 1.416e-01 3.346e-02 4.233 2.45e-05 ***
educ3 3.087e-01 3.227e-02 9.567 < 2e-16 ***
educ4 4.812e-01 3.332e-02 14.442 < 2e-16 ***
educ5-élevé 6.407e-01 3.343e-02 19.166 < 2e-16 ***
exper 3.437e-02 2.539e-03 13.535 < 2e-16 ***
I(exper^2) -4.928e-04 6.398e-05 -7.702 2.45e-14 ***
sexHomme 1.152e-01 1.545e-02 7.457 1.51e-13 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.2816 on 1464 degrees of freedom

Multiple R-squared: 0.3995, Adjusted R-squared: 0.3966
F-statistic: 139.1 on 7 and 1464 DF, p-value: < 2.2e-16

>
> # Modèle 3
>
> lm.fit3 <- lm(f3, data = Bwages, na.action = na.exclude)
> summary(lm.fit3)

Call:
lm(formula = f3, data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.53173 -0.15037 0.00789 0.16645 1.07623

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.067e+00 6.299e-02 80.440 < 2e-16 ***
educ2 2.156e-01 6.762e-02 3.189 0.00146 **
educ3 4.456e-01 6.298e-02 7.076 2.30e-12 ***
educ4 6.190e-01 6.247e-02 9.908 < 2e-16 ***
educ5-élevé 7.719e-01 6.404e-02 12.053 < 2e-16 ***
exper 3.465e-02 2.539e-03 13.648 < 2e-16 ***
I(exper^2) -4.979e-04 6.391e-05 -7.790 1.26e-14 ***
sexHomme 2.741e-01 6.691e-02 4.096 4.43e-05 ***
educ2:sexHomme -9.324e-02 7.756e-02 -1.202 0.22951
educ3:sexHomme -1.821e-01 7.256e-02 -2.509 0.01220 *
educ4:sexHomme -1.912e-01 7.324e-02 -2.610 0.00914 **
educ5-élevé:sexHomme -1.726e-01 7.393e-02 -2.335 0.01970 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.281 on 1460 degrees of freedom

Multiple R-squared: 0.4038, Adjusted R-squared: 0.3993
F-statistic: 89.88 on 11 and 1460 DF, p-value: < 2.2e-16

>
> # Modèle 4
>
> lm.fit4 <- lm(f4, data = Bwages, na.action = na.exclude)
> summary(lm.fit4)

Call:
lm(formula = f4, data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.43944 -0.15304 0.00623 0.16474 1.09383

Coefficients:

Page 3 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

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

(Intercept) 5.120e+00 6.666e-02 76.802 < 2e-16 ***
educ2 2.095e-01 6.756e-02 3.100 0.00197 **
educ3 4.242e-01 6.351e-02 6.679 3.41e-11 ***
educ4 5.947e-01 6.319e-02 9.412 < 2e-16 ***
educ5-élevé 7.450e-01 6.493e-02 11.475 < 2e-16 ***
exper 3.301e-02 2.627e-03 12.567 < 2e-16 ***
I(exper^2) -5.189e-04 6.442e-05 -8.055 1.63e-15 ***
sexHomme 1.797e-01 7.766e-02 2.314 0.02080 *
educ2:sexHomme -7.907e-02 7.767e-02 -1.018 0.30882
educ3:sexHomme -1.498e-01 7.370e-02 -2.033 0.04225 *
educ4:sexHomme -1.531e-01 7.485e-02 -2.046 0.04095 *
educ5-élevé:sexHomme -1.313e-01 7.581e-02 -1.732 0.08345 .
exper:sexHomme 3.826e-03 1.605e-03 2.384 0.01726 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.2805 on 1459 degrees of freedom

Multiple R-squared: 0.4061, Adjusted R-squared: 0.4012
F-statistic: 83.13 on 12 and 1459 DF, p-value: < 2.2e-16

>
> # b) Statistiques AIC et BIC
>
> AIC(lm.fit0, lm.fit1, lm.fit2, lm.fit3, lm.fit4)
df AIC
lm.fit0 2 1193.2707
lm.fit1 8 509.4604
lm.fit2 9 456.5910
lm.fit3 13 454.0713
lm.fit4 14 450.3493

> BIC(lm.fit0, lm.fit1, lm.fit2, lm.fit3, lm.fit4)

df BIC
lm.fit0 2 1203.8594
lm.fit1 8 551.8154
lm.fit2 9 504.2404
lm.fit3 13 522.8982
lm.fit4 14 524.4706

>
> # c) Tests de modèles
>
> # Tests de significativité des coefficients du modèle 1
>
> coeftest(lm.fit1) # t-test

t test of coefficients:

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

(Intercept) 5.2623e+00 3.8085e-02 138.1727 < 2.2e-16 ***
educ2 1.4117e-01 3.4080e-02 4.1424 3.633e-05 ***
educ3 2.9670e-01 3.2825e-02 9.0387 < 2.2e-16 ***
educ4 4.5290e-01 3.3715e-02 13.4329 < 2.2e-16 ***
educ5-élevé 6.2903e-01 3.4010e-02 18.4957 < 2.2e-16 ***
exper 3.5298e-02 2.5828e-03 13.6663 < 2.2e-16 ***
I(exper^2) -4.9989e-04 6.5155e-05 -7.6724 3.061e-14 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

>
> # Tests d'ajustement général du modèle 1
>
> waldtest(lm.fit1, test = "F")

Page 4 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

Wald test

Model 1: log(wage) ~ educ + exper + I(exper^2)

Model 2: log(wage) ~ 1
Res.Df Df F Pr(>F)
1 1465
2 1471 -6 147.55 < 2.2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
> # Test du modèle 4 contre le modèle 1
>
> anova(lm.fit1, lm.fit4)
Analysis of Variance Table

Model 1: log(wage) ~ educ + exper + I(exper^2)

Model 2: log(wage) ~ educ + exper + I(exper^2) + sex + educ:sex + exper:sex
Res.Df RSS Df Sum of Sq F Pr(>F)
1 1465 120.51
2 1459 114.83 6 5.6833 12.036 3.11e-13 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
> # ou...
>
> waldtest(lm.fit1, lm.fit4, test = "F")
Wald test

Model 1: log(wage) ~ educ + exper + I(exper^2)

Model 2: log(wage) ~ educ + exper + I(exper^2) + sex + educ:sex + exper:sex
Res.Df Df F Pr(>F)
1 1465
2 1459 6 12.036 3.11e-13 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
> # d) Présentation des modèles
>
>
> mtable(lm.fit1, lm.fit2, lm.fit3, lm.fit4,
+ summary.stats = c("R-squared", "adj. R-squared", "F", "AIC", "N"))

Calls:
lm.fit1: lm(formula = f1, data = Bwages, na.action = na.exclude)
lm.fit2: lm(formula = f2, data = Bwages, na.action = na.exclude)
lm.fit3: lm(formula = f3, data = Bwages, na.action = na.exclude)
lm.fit4: lm(formula = f4, data = Bwages, na.action = na.exclude)

======================================================================================
lm.fit1 lm.fit2 lm.fit3 lm.fit4
--------------------------------------------------------------------------------------
(Intercept) 5.262*** 5.193*** 5.067*** 5.120***
(0.038) (0.039) (0.063) (0.067)
educ: 2/1-bas 0.141*** 0.142*** 0.216** 0.209**
(0.034) (0.033) (0.068) (0.068)
educ: 3/1-bas 0.297*** 0.309*** 0.446*** 0.424***
(0.033) (0.032) (0.063) (0.064)
educ: 4/1-bas 0.453*** 0.481*** 0.619*** 0.595***
(0.034) (0.033) (0.062) (0.063)
educ: 5-élevé/1-bas 0.629*** 0.641*** 0.772*** 0.745***
(0.034) (0.033) (0.064) (0.065)
exper 0.035*** 0.034*** 0.035*** 0.033***
(0.003) (0.003) (0.003) (0.003)
I(exper^2) -0.000*** -0.000*** -0.000*** -0.001***
(0.000) (0.000) (0.000) (0.000)
sex: Homme/Femme 0.115*** 0.274*** 0.180*

Page 5 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

(0.015) (0.067) (0.078)

educ: 2/1-bas x sexHomme -0.093 -0.079
(0.078) (0.078)
educ: 3/1-bas x sexHomme -0.182* -0.150*
(0.073) (0.074)
educ: 4/1-bas x sexHomme -0.191** -0.153*
(0.073) (0.075)
educ: 5-élevé/1-bas x sexHomme -0.173* -0.131
(0.074) (0.076)
exper x sexHomme 0.004*
(0.002)
--------------------------------------------------------------------------------------
R-squared 0.377 0.399 0.404 0.406
adj. R-squared 0.374 0.397 0.399 0.401
F 147.554 139.131 89.881 83.129
AIC 509.460 456.591 454.071 450.349
N 1472 1472 1472 1472
======================================================================================
Significance: *** = p < 0.001; ** = p < 0.01; * = p < 0.05
>
>
> # 3. Sélection de modèles
> # -----------------------
>
> # a) Sélection forward
>
> stepreg.fit1 <- step(lm.fit0,
+ scope = ~ I(exper^2) + sex*educ + sex*exper,
+ direction = "forward", trace = T)
Start: AIC=-2986.08
log(wage) ~ 1

Df Sum of Sq RSS AIC

+ educ 4 31.1125 162.22 -3236.4
+ exper 1 18.3192 175.02 -3130.6
+ I(exper^2) 1 12.1568 181.18 -3079.7
+ sex 1 4.3744 188.96 -3017.8
<none> 193.34 -2986.1

Step: AIC=-3236.35
log(wage) ~ educ

Df Sum of Sq RSS AIC

+ exper 1 36.871 125.35 -3613.9
+ I(exper^2) 1 26.350 135.87 -3495.3
+ sex 1 7.960 154.26 -3308.4
<none> 162.22 -3236.4

Step: AIC=-3613.9
log(wage) ~ educ + exper

Df Sum of Sq RSS AIC

+ I(exper^2) 1 4.8423 120.51 -3669.9
+ sex 1 4.5472 120.81 -3666.3
<none> 125.35 -3613.9

Step: AIC=-3669.89
log(wage) ~ educ + exper + I(exper^2)

Df Sum of Sq RSS AIC

+ sex 1 4.4094 116.10 -3722.8
<none> 120.51 -3669.9

Step: AIC=-3722.76
log(wage) ~ educ + exper + I(exper^2) + sex

Page 6 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

Df Sum of Sq RSS AIC

+ sex:exper 1 0.79432 115.31 -3730.9
+ sex:educ 4 0.82675 115.27 -3725.3
<none> 116.10 -3722.8

Step: AIC=-3730.87
log(wage) ~ educ + exper + I(exper^2) + sex + exper:sex

Df Sum of Sq RSS AIC

<none> 115.31 -3730.9
+ sex:educ 4 0.47967 114.83 -3729.0
> summary(stepreg.fit1)

Call:
lm(formula = log(wage) ~ educ + exper + I(exper^2) + sex + exper:sex,
data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.41148 -0.15152 0.00854 0.16322 1.11280

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.228e+00 3.998e-02 130.752 < 2e-16 ***
educ2 1.484e-01 3.343e-02 4.440 9.65e-06 ***
educ3 3.131e-01 3.220e-02 9.723 < 2e-16 ***
educ4 4.853e-01 3.324e-02 14.598 < 2e-16 ***
educ5-élevé 6.451e-01 3.335e-02 19.340 < 2e-16 ***
exper 3.239e-02 2.607e-03 12.428 < 2e-16 ***
I(exper^2) -5.216e-04 6.443e-05 -8.096 1.18e-15 ***
sexHomme 3.515e-02 2.956e-02 1.189 0.23462
exper:sexHomme 4.864e-03 1.532e-03 3.175 0.00153 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.2807 on 1463 degrees of freedom

Multiple R-squared: 0.4036, Adjusted R-squared: 0.4003
F-statistic: 123.8 on 8 and 1463 DF, p-value: < 2.2e-16

>
> # b) Sélection backward
>
> stepreg.fit2 <- step(lm.fit4,
+ scope = list(upper =~., lower = ~1.),
+ direction = "backward", trace=T)
Start: AIC=-3729.01
log(wage) ~ educ + exper + I(exper^2) + sex + educ:sex + exper:sex

Df Sum of Sq RSS AIC

- educ:sex 4 0.4797 115.31 -3730.9
<none> 114.83 -3729.0
- exper:sex 1 0.4472 115.27 -3725.3
- I(exper^2) 1 5.1069 119.93 -3667.0

Step: AIC=-3730.87
log(wage) ~ educ + exper + I(exper^2) + sex + exper:sex

Df Sum of Sq RSS AIC

<none> 115.31 -3730.9
- exper:sex 1 0.794 116.10 -3722.8
- I(exper^2) 1 5.166 120.47 -3668.4
- educ 4 52.340 167.65 -3187.9
> summary(stepreg.fit2)

Page 7 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

Call:
lm(formula = log(wage) ~ educ + exper + I(exper^2) + sex + exper:sex,
data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.41148 -0.15152 0.00854 0.16322 1.11280

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.228e+00 3.998e-02 130.752 < 2e-16 ***
educ2 1.484e-01 3.343e-02 4.440 9.65e-06 ***
educ3 3.131e-01 3.220e-02 9.723 < 2e-16 ***
educ4 4.853e-01 3.324e-02 14.598 < 2e-16 ***
educ5-élevé 6.451e-01 3.335e-02 19.340 < 2e-16 ***
exper 3.239e-02 2.607e-03 12.428 < 2e-16 ***
I(exper^2) -5.216e-04 6.443e-05 -8.096 1.18e-15 ***
sexHomme 3.515e-02 2.956e-02 1.189 0.23462
exper:sexHomme 4.864e-03 1.532e-03 3.175 0.00153 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.2807 on 1463 degrees of freedom

Multiple R-squared: 0.4036, Adjusted R-squared: 0.4003
F-statistic: 123.8 on 8 and 1463 DF, p-value: < 2.2e-16

>
> # c) Sélection forward-backward
>
> stepreg.fit3 <- step(lm.fit0,
+ scope = ~ I(exper^2) + sex*educ + sex*exper,
+ direction = "both", trace=T)
Start: AIC=-2986.08
log(wage) ~ 1

Df Sum of Sq RSS AIC

+ educ 4 31.1125 162.22 -3236.4
+ exper 1 18.3192 175.02 -3130.6
+ I(exper^2) 1 12.1568 181.18 -3079.7
+ sex 1 4.3744 188.96 -3017.8
<none> 193.34 -2986.1

Step: AIC=-3236.35
log(wage) ~ educ

Df Sum of Sq RSS AIC

+ exper 1 36.871 125.35 -3613.9
+ I(exper^2) 1 26.350 135.87 -3495.3
+ sex 1 7.960 154.26 -3308.4
<none> 162.22 -3236.4
- educ 4 31.112 193.34 -2986.1

Step: AIC=-3613.9
log(wage) ~ educ + exper

Df Sum of Sq RSS AIC

+ I(exper^2) 1 4.842 120.51 -3669.9
+ sex 1 4.547 120.81 -3666.3
<none> 125.35 -3613.9
- exper 1 36.871 162.22 -3236.4
- educ 4 49.665 175.02 -3130.6

Step: AIC=-3669.89
log(wage) ~ educ + exper + I(exper^2)

Page 8 sur 9
Exemples-Sélection de modèles 04.11.19 07:45

Df Sum of Sq RSS AIC

+ sex 1 4.409 116.10 -3722.8
<none> 120.51 -3669.9
- I(exper^2) 1 4.842 125.35 -3613.9
- exper 1 15.363 135.87 -3495.3
- educ 4 49.899 170.41 -3167.9

Step: AIC=-3722.76
log(wage) ~ educ + exper + I(exper^2) + sex

Df Sum of Sq RSS AIC

+ sex:exper 1 0.794 115.31 -3730.9
+ sex:educ 4 0.827 115.27 -3725.3
<none> 116.10 -3722.8
- sex 1 4.409 120.51 -3669.9
- I(exper^2) 1 4.704 120.81 -3666.3
- exper 1 14.529 130.63 -3551.2
- educ 4 52.347 168.45 -3182.9

Step: AIC=-3730.87
log(wage) ~ educ + exper + I(exper^2) + sex + exper:sex

Df Sum of Sq RSS AIC

<none> 115.31 -3730.9
+ sex:educ 4 0.480 114.83 -3729.0
- exper:sex 1 0.794 116.10 -3722.8
- I(exper^2) 1 5.166 120.47 -3668.4
- educ 4 52.340 167.65 -3187.9
> summary(stepreg.fit3)

Call:
lm(formula = log(wage) ~ educ + exper + I(exper^2) + sex + exper:sex,
data = Bwages, na.action = na.exclude)

Residuals:
Min 1Q Median 3Q Max
-1.41148 -0.15152 0.00854 0.16322 1.11280

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.228e+00 3.998e-02 130.752 < 2e-16 ***
educ2 1.484e-01 3.343e-02 4.440 9.65e-06 ***
educ3 3.131e-01 3.220e-02 9.723 < 2e-16 ***
educ4 4.853e-01 3.324e-02 14.598 < 2e-16 ***
educ5-élevé 6.451e-01 3.335e-02 19.340 < 2e-16 ***
exper 3.239e-02 2.607e-03 12.428 < 2e-16 ***
I(exper^2) -5.216e-04 6.443e-05 -8.096 1.18e-15 ***
sexHomme 3.515e-02 2.956e-02 1.189 0.23462
exper:sexHomme 4.864e-03 1.532e-03 3.175 0.00153 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.2807 on 1463 degrees of freedom

Multiple R-squared: 0.4036, Adjusted R-squared: 0.4003
F-statistic: 123.8 on 8 and 1463 DF, p-value: < 2.2e-16

>
>

Page 9 sur 9