1. Importer les données

donnees <- read.table("https://r-stat-sc-donnees.github.io/ronfle.txt",header=TRUE)
summary(donnees)

      age            poids            taille          alcool      sexe   ronfle taba  
 Min.   :23.00   Min.   : 42.00   Min.   :158.0   Min.   : 0.00   F:25   N:65   N:36  
 1st Qu.:43.00   1st Qu.: 77.00   1st Qu.:166.0   1st Qu.: 0.00   H:75   O:35   O:64  
 Median :52.00   Median : 95.00   Median :186.0   Median : 2.00                       
 Mean   :52.27   Mean   : 90.41   Mean   :181.1   Mean   : 2.95                       
 3rd Qu.:62.25   3rd Qu.:107.00   3rd Qu.:194.0   3rd Qu.: 4.25                       
 Max.   :74.00   Max.   :120.00   Max.   :208.0   Max.   :15.00                       

2. Construire le modèle

library(MASS)
mod.complet <- lda(ronfle~.,data=donnees)
mod.complet

Call:
lda(ronfle ~ ., data = donnees)

Prior probabilities of groups:
   N    O 
0.65 0.35 

Group means:
       age    poids   taille   alcool     sexeH     tabaO
N 50.26154 90.47692 180.9538 2.369231 0.6923077 0.6769231
O 56.00000 90.28571 181.3714 4.028571 0.8571429 0.5714286

Coefficients of linear discriminants:
               LD1
age     0.05973655
poids  -0.01620579
taille  0.01590170
alcool  0.24058822
sexeH   0.55413371
tabaO  -1.14621434

plot(mod.complet)

3. Sélectionner des variables

mat.X <- model.matrix(ronfle~.,data=donnees)[,-1]
head(mat.X)

  age poids taille alcool sexeH tabaO
1  47    71    158      0     1     1
2  56    58    164      7     1     0
3  46   116    208      3     1     1
4  70    96    186      3     1     1
5  51    91    195      2     1     1
6  46    98    188      0     0     0

library(klaR)
ronfle <- donnees$ronfle
set.seed(1234)
sel <- stepclass(mat.X,donnees$ronfle,method="lda",direction="both",maxvar=6)

correctness rate: 0.65;  in: "poids";  variables (1): poids 
correctness rate: 0.65;  in: "age";  variables (2): poids, age 
correctness rate: 0.68;  in: "alcool";  variables (3): poids, age, alcool 

 hr.elapsed min.elapsed sec.elapsed 
       0.00        0.00        0.89 

mod.sel <- lda(sel$formula,data=donnees)
mod.sel

Call:
lda(sel$formula, data = donnees)

Prior probabilities of groups:
   N    O 
0.65 0.35 

Group means:
       age    poids   alcool
N 50.26154 90.47692 2.369231
O 56.00000 90.28571 4.028571

Coefficients of linear discriminants:
                LD1
age     0.072815712
poids  -0.004607728
alcool  0.245088681

4. Estimer le taux de mauvais classement

prev.complet <- lda(ronfle~.,data=donnees,CV=TRUE)$class
prev.sel <- lda(sel$formula,data=donnees,CV=TRUE)$class

table(prev.complet,donnees$ronfle)

            
prev.complet  N  O
           N 53 22
           O 12 13

table(prev.sel,donnees$ronfle)

        
prev.sel  N  O
       N 55 23
       O 10 12

mean(prev.complet!=donnees$ronfle)

[1] 0.34

mean(prev.sel!=donnees$ronfle)

[1] 0.33

5. Faire de la prévision

n_don1 <- matrix(c(42,55,169,0,58,94,185,4,35,70,180,6,67,63,166,3),ncol=4,byrow=T)
n_don2 <- matrix(c("F","N","H","O","H","O","F","N"),ncol=2,byrow=T)
n_donnees <- cbind.data.frame(n_don1,n_don2)
names(n_donnees) <- names(donnees)[-6]
predict(mod.sel,newdata=n_donnees)

$`class`
[1] N N N O
Levels: N O

$posterior
          N         O
1 0.8582230 0.1417770
2 0.5444034 0.4555966
3 0.7435646 0.2564354
4 0.4308341 0.5691659

$x
         LD1
1 -1.3076693
2  0.6580354
3 -0.4159631
4  1.2111277