Analyses Article 2

1 Donnees

On a à la base 84 T et 85 BL (on a enlevé les 22 intermédiaires), dont respectivement 45 T et 44 BL ont les résultats du multimorphe. On avait donc 89 personnes ayant répondu au MM.

73 ont répondu au MM et avaient moins de 4 valeurs manquantes sur les autres variables nécessaires pour les modèles. C’est à dire les variables composant le score abandon :

PBI_soin_Mere , PBI_soin_Pere , CTQ_EN , CTQ_PN

et les variables composant le score intrusion :

PBI_control Pere , PBI control mere , CTQ_EA , CTQ_SA , CTQ_PA

Les scores Abandon et Intrusion utilisés ici ne sont pas les simples sommes , mais bien les première dimensions des ACPs faites à partir des “paniers” de variables présentés ci dessus. LEs résultats de ces ACP sont dans l’autre document Rpub en ligne.

1.1 CSP

on a importé les CSP données par Marion : 3 classes : 1(professions supérieures) , 2 (profession intermédiaires), 3 (retraité et chomage) On avait seulement 8 personnes dans la 3eme catégorie, on l’a regroupé avec la 2.

1.2 correlations

Table 1.1: Correlations sur echantillon Total

variable 1	variable 2	coeff de correlation	p-value	N
MM_Delai	Dim.Abandon	0.14	0.2545	70
MM_Delai	Dim.Intrusion	0.21	0.0779	70
MM_Succes	Dim.Abandon	-0.028	0.8201	70
MM_Succes	Dim.Intrusion	-0.42	0.0003	70

Table 1.1: Correlations sur BL

variable 1	variable 2	coeff de correlation	p-value	N
MM_Delai	Dim.Abandon	-0.27	0.1722	27
MM_Delai	Dim.Intrusion	0.22	0.2780	26
MM_Succes	Dim.Abandon	-0.0038	0.9849	27
MM_Succes	Dim.Intrusion	-0.69	0.0001	26

Table 1.1: Correlations sur T

variable 1	variable 2	coeff de correlation	p-value	N
MM_Delai	Dim.Abandon	0.31	0.0402	43
MM_Delai	Dim.Intrusion	0.055	0.7236	44
MM_Succes	Dim.Abandon	0.085	0.5859	43
MM_Succes	Dim.Intrusion	-0.023	0.8822	44

2 MM et abandon/intrusion

2.1 ACPs

Abandon puis Intrusion :

2.2 Modeles

CHOIX FINAL : Evt_TS dans la Dim.intrusion

Voici le modèle final (modèle de régression linéaire ) modélisant le % de Succès au MM :

Call:
lm(formula = MM_Succes ~ age + sexe + CSP_bin2 + Dim.Intrusion + 
    Dim.Abandon * borderline, data = DBsi)

Residuals:
     Min       1Q   Median       3Q      Max 
-0.11564 -0.06130  0.01135  0.04962  0.10524 

Coefficients:
                        Estimate Std. Error t value Pr(>|t|)    
(Intercept)             0.461248   0.125034   3.689 0.000493 ***
age                     0.020635   0.008018   2.574 0.012600 *  
sexe                    0.060519   0.024739   2.446 0.017433 *  
CSP_bin2                0.011803   0.019666   0.600 0.550707    
Dim.Intrusion          -0.031146   0.006262  -4.974    6e-06 ***
Dim.Abandon             0.029970   0.010188   2.942 0.004657 ** 
borderline             -0.014169   0.022228  -0.637 0.526299    
Dim.Abandon:borderline -0.031158   0.013836  -2.252 0.028063 *  
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.06507 on 59 degrees of freedom
  (22 observations deleted due to missingness)
Multiple R-squared:  0.4367,    Adjusted R-squared:  0.3699 
F-statistic: 6.535 on 7 and 59 DF,  p-value: 1.018e-05

Voici le modèle final (modèle de régression linéaire ) modélisant le Délai au MM :

Call:
lm(formula = MM_Delai ~ age + sexe + CSP_bin2 + Dim.Intrusion + 
    Dim.Abandon * borderline, data = DBsi)

Residuals:
     Min       1Q   Median       3Q      Max 
-10.1771  -2.9367  -0.0382   3.2707  11.6718 

Coefficients:
                       Estimate Std. Error t value Pr(>|t|)    
(Intercept)            37.90134    9.36091   4.049 0.000152 ***
age                    -0.50964    0.60031  -0.849 0.399332    
sexe                   -1.83471    1.85214  -0.991 0.325931    
CSP_bin2                4.52643    1.47236   3.074 0.003194 ** 
Dim.Intrusion           0.02866    0.46882   0.061 0.951456    
Dim.Abandon             1.20010    0.76272   1.573 0.120964    
borderline              2.96752    1.66414   1.783 0.079695 .  
Dim.Abandon:borderline -2.23515    1.03587  -2.158 0.035029 *  
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 4.872 on 59 degrees of freedom
  (22 observations deleted due to missingness)
Multiple R-squared:  0.2655,    Adjusted R-squared:  0.1783 
F-statistic: 3.046 on 7 and 59 DF,  p-value: 0.008358

3 REMARQUES !!

• L’imputation ne change pas les résultats du modèle MM_Succes, et deteriore même un peu la qualité du modèle MM_Delai. Je suis donc d’avis qu’on ne fasse pas d’imputation, cela enlève un angle d’attaque sur notre article.

• EVt_TS : d’un point de vue statistique, il a y + de cohérence à le mettre dans le score Intrusion. Les modèles ne sont que renfocrés dans leur qualités, et les effets sont légèrement + significatifs.

3.1 modèles intermédiaires

Call:
lm(formula = MM_Succes ~ Dim.Intrusion + Dim.Abandon, data = DBsi)

Residuals:
      Min        1Q    Median        3Q       Max 
-0.193875 -0.057218  0.002809  0.064022  0.126149 

Coefficients:
               Estimate Std. Error t value Pr(>|t|)    
(Intercept)    0.832027   0.008738  95.217  < 2e-16 ***
Dim.Intrusion -0.027411   0.006413  -4.275  6.3e-05 ***
Dim.Abandon    0.011876   0.006637   1.790   0.0781 .  
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.07258 on 66 degrees of freedom
  (20 observations deleted due to missingness)
Multiple R-squared:  0.2174,    Adjusted R-squared:  0.1937 
F-statistic: 9.169 on 2 and 66 DF,  p-value: 0.0003063

Call:
lm(formula = MM_Succes ~ age + sexe + CSP_bin2 + Dim.Intrusion + 
    Dim.Abandon, data = DBsi)

Residuals:
      Min        1Q    Median        3Q       Max 
-0.150471 -0.049074  0.008872  0.056280  0.101926 

Coefficients:
               Estimate Std. Error t value Pr(>|t|)    
(Intercept)    0.453972   0.128064   3.545 0.000761 ***
age            0.020806   0.008193   2.540 0.013663 *  
sexe           0.043119   0.024244   1.779 0.080305 .  
CSP_bin2       0.013484   0.020271   0.665 0.508445    
Dim.Intrusion -0.028939   0.005994  -4.828 9.63e-06 ***
Dim.Abandon    0.009682   0.006215   1.558 0.124438    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.06722 on 61 degrees of freedom
  (22 observations deleted due to missingness)
Multiple R-squared:  0.3786,    Adjusted R-squared:  0.3276 
F-statistic: 7.432 on 5 and 61 DF,  p-value: 1.69e-05

Call:
lm(formula = MM_Succes ~ age + sexe + CSP_bin2 + Dim.Intrusion + 
    Dim.Abandon + borderline, data = DBsi)

Residuals:
     Min       1Q   Median       3Q      Max 
-0.15341 -0.05390  0.01181  0.05257  0.10688 

Coefficients:
               Estimate Std. Error t value Pr(>|t|)    
(Intercept)    0.440442   0.128853   3.418  0.00114 ** 
age            0.021864   0.008267   2.645  0.01042 *  
sexe           0.049164   0.025028   1.964  0.05412 .  
CSP_bin2       0.012184   0.020322   0.600  0.55105    
Dim.Intrusion -0.027301   0.006226  -4.385 4.75e-05 ***
Dim.Abandon    0.012898   0.007033   1.834  0.07161 .  
borderline    -0.022180   0.022674  -0.978  0.33188    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.06725 on 60 degrees of freedom
  (22 observations deleted due to missingness)
Multiple R-squared:  0.3883,    Adjusted R-squared:  0.3272 
F-statistic: 6.349 on 6 and 60 DF,  p-value: 3.261e-05

3.2 modèles avec interactions borderlines pour Abandon et Intrusion

% de Succès au MM :

Call:
lm(formula = MM_Succes ~ age + sexe + CSP_bin2 + Dim.Intrusion * 
    borderline + Dim.Abandon * borderline, data = DBsi)

Residuals:
      Min        1Q    Median        3Q       Max 
-0.122878 -0.042461  0.004129  0.053541  0.116145 

Coefficients:
                          Estimate Std. Error t value Pr(>|t|)    
(Intercept)               0.468566   0.120378   3.892 0.000259 ***
age                       0.020152   0.007720   2.610 0.011489 *  
sexe                      0.066129   0.023926   2.764 0.007642 ** 
CSP_bin2                  0.007400   0.019018   0.389 0.698619    
Dim.Intrusion            -0.007235   0.011694  -0.619 0.538538    
borderline               -0.009731   0.021474  -0.453 0.652119    
Dim.Abandon               0.015593   0.011508   1.355 0.180681    
Dim.Intrusion:borderline -0.032523   0.013630  -2.386 0.020309 *  
borderline:Dim.Abandon   -0.016908   0.014594  -1.159 0.251388    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.06263 on 58 degrees of freedom
  (22 observations deleted due to missingness)
Multiple R-squared:  0.4871,    Adjusted R-squared:  0.4164 
F-statistic: 6.885 on 8 and 58 DF,  p-value: 2.487e-06

Délai moyen au MM :

Call:
lm(formula = MM_Delai ~ age + sexe + CSP_bin2 + Dim.Intrusion * 
    borderline + Dim.Abandon * borderline, data = DBsi)

Residuals:
     Min       1Q   Median       3Q      Max 
-10.9738  -2.6994  -0.2573   2.4100  11.3481 

Coefficients:
                         Estimate Std. Error t value Pr(>|t|)    
(Intercept)               37.3667     9.0334   4.136 0.000115 ***
age                       -0.4744     0.5793  -0.819 0.416244    
sexe                      -2.2446     1.7955  -1.250 0.216273    
CSP_bin2                   4.8481     1.4271   3.397 0.001236 ** 
Dim.Intrusion             -1.7182     0.8775  -1.958 0.055044 .  
borderline                 2.6433     1.6115   1.640 0.106351    
Dim.Abandon                2.2504     0.8636   2.606 0.011629 *  
Dim.Intrusion:borderline   2.3760     1.0228   2.323 0.023708 *  
borderline:Dim.Abandon    -3.2762     1.0952  -2.991 0.004072 ** 
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 4.7 on 58 degrees of freedom
  (22 observations deleted due to missingness)
Multiple R-squared:  0.328, Adjusted R-squared:  0.2353 
F-statistic: 3.539 on 8 and 58 DF,  p-value: 0.002128

3.3 modèles avec polynomes orthogonaux

% de Succès au MM :

Call:
lm(formula = MM_Succes ~ age + sexe + CSP_bin2 + poly(Dim.Intrusion, 
    2) * borderline + poly(Dim.Abandon, 2) * borderline, data = DBsi[which(!is.na(DBsi$Dim.Abandon) & 
    !is.na(DBsi$Dim.Intrusion)), ])

Residuals:
      Min        1Q    Median        3Q       Max 
-0.127233 -0.038965  0.001864  0.050848  0.116081 

Coefficients:
                                    Estimate Std. Error t value Pr(>|t|)    
(Intercept)                         0.459735   0.129955   3.538 0.000838 ***
age                                 0.020607   0.008297   2.484 0.016143 *  
sexe                                0.065940   0.025431   2.593 0.012221 *  
CSP_bin2                            0.004134   0.021025   0.197 0.844877    
poly(Dim.Intrusion, 2)1            -0.126327   0.320107  -0.395 0.694663    
poly(Dim.Intrusion, 2)2            -0.067786   0.307890  -0.220 0.826574    
borderline                         -0.006151   0.031208  -0.197 0.844496    
poly(Dim.Abandon, 2)1               0.133211   0.228736   0.582 0.562735    
poly(Dim.Abandon, 2)2              -0.058790   0.171976  -0.342 0.733792    
poly(Dim.Intrusion, 2)1:borderline -0.363580   0.343668  -1.058 0.294793    
poly(Dim.Intrusion, 2)2:borderline  0.045266   0.326387   0.139 0.890211    
borderline:poly(Dim.Abandon, 2)1   -0.186034   0.301465  -0.617 0.539761    
borderline:poly(Dim.Abandon, 2)2    0.090985   0.223995   0.406 0.686207    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.06469 on 54 degrees of freedom
  (2 observations deleted due to missingness)
Multiple R-squared:  0.4905,    Adjusted R-squared:  0.3773 
F-statistic: 4.332 on 12 and 54 DF,  p-value: 8.601e-05

Délai moyen au MM :

Call:
lm(formula = MM_Delai ~ age + sexe + CSP_bin2 + poly(Dim.Intrusion, 
    2) * borderline + poly(Dim.Abandon, 2) * borderline, data = DBsi[which(!is.na(DBsi$Dim.Abandon) & 
    !is.na(DBsi$Dim.Intrusion)), ])

Residuals:
    Min      1Q  Median      3Q     Max 
-9.8853 -2.7825 -0.1436  2.4266 11.1162 

Coefficients:
                                   Estimate Std. Error t value Pr(>|t|)    
(Intercept)                         40.3853     9.0891   4.443 4.44e-05 ***
age                                 -0.7156     0.5803  -1.233 0.222859    
sexe                                -1.4869     1.7787  -0.836 0.406850    
CSP_bin2                             5.4886     1.4705   3.733 0.000457 ***
poly(Dim.Intrusion, 2)1            -52.3065    22.3885  -2.336 0.023216 *  
poly(Dim.Intrusion, 2)2            -32.9893    21.5340  -1.532 0.131370    
borderline                           6.4017     2.1827   2.933 0.004919 ** 
poly(Dim.Abandon, 2)1               36.5236    15.9980   2.283 0.026391 *  
poly(Dim.Abandon, 2)2                8.1343    12.0281   0.676 0.501757    
poly(Dim.Intrusion, 2)1:borderline  52.0757    24.0364   2.167 0.034702 *  
poly(Dim.Intrusion, 2)2:borderline  45.6602    22.8277   2.000 0.050517 .  
borderline:poly(Dim.Abandon, 2)1   -71.4246    21.0847  -3.388 0.001323 ** 
borderline:poly(Dim.Abandon, 2)2    14.0350    15.6664   0.896 0.374301    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 4.525 on 54 degrees of freedom
  (2 observations deleted due to missingness)
Multiple R-squared:  0.4201,    Adjusted R-squared:  0.2913 
F-statistic: 3.261 on 12 and 54 DF,  p-value: 0.001395