Je suis entrain de réaliser un travail assez lourd pour un novice comme moi qui consiste en la détermination du seuil de résistance d'un micro organisme à une molécule utilisée en traitement.
Dans un premier temps je veux observer la distribution de mes échantillons, qui n'est pas Gaussienne et probablement multicomposante (les mesures sont les concentrations inhibitrice pour lesquelles 50% des micro organismes meurent), pour cela j'ai utilisé la fonction mixmodCluster du package Rmixmod (https://cran.r-project.org/web/packages ... index.html)
Voici mon script utilisé :
Code : Tout sélectionner
BM<-c(9.19, 0.5, 10.61, 7.98, 10.34, 6.03, 14.6, 39.82, 8.07, 9.33, 17.57, 7.14, 21.92, 8.68, 3.05, 8.65, 7.84, 8.89, 7.44, 6.68, 21.98, 4.16, 12.54, 4.84, 23.4, 9.97, 11.4, 11.56, 31.59, 5.71, 5.31, 14.83, 2.82, 8.8, 6.42, 9.72, 24.5, 4.65, 14.93, 9.3, 9.52, 4.38, 8.53, 12.46, 6.19, 5.46, 39.62, 9.25, 4.5, 12.37, 6.95, 1.38, 7.11, 34, 11.96, 5.18, 4.61, 17.69, 14.86, 7.68, 12.19, 4.18, 1.87, 6.9, 56.2, 84.2, 7.48, 4.57, 2.52, 4.5, 11.8, 7.7, 6.78, 65.9, 5.9, 87.2, 18.6, 12.3, 3.3, 26.6, 4.3, 0.69, 7.6, 14.2, 7.1, 17.2, 4.3, 19.9, 9.8, 17.2, 3.8, 16.42, 1.94, 5.2, 20.1, 6.5, 9.7, 4.2, 3.76, 15.8, 18.46, 9.81, 1.72, 2.29, 9.51, 10.14, 11.7, 8.68, 9.84, 1.88, 10.98, 6.3, 6.6, 6.76, 6.45, 1.6, 4, 1.06, 2.73, 8.67, 14.4, 7.38, 4.71, 1.86, 6, 4.1, 7.87, 2.64, 9.2, 2.19, 6.7, 0.87, 3.33, 8.72, 0.167, 13.57, 11.9, 5.39, 13.64, 22.67, 11.11, 7.96, 0.35, 2.75, 8.31, 9.19, 2.75, 4.33, 10.88, 18.2, 8.59, 11.37, 4.1, 9.72, 6.42, 3.04, 11, 2.22, 2.64, 2.86, 13.8, 18.2, 9.42, 23.7, 8.02, 3.75, 6.79, 16.63, 19.86, 17.63, 14.25, 11.98, 6.72, 3.73, 4.37, 11.48, 12.9, 0.4, 3.49, 1.06, 8.94, 12.1, 7.88, 4, 0.64, 1.51, 9.34, 4.98, 8.36, 2.14, 32.45, 4.7, 4.7, 5.65, 1.84, 1.87, 8.5, 21.43, 2.67, 0.73, 3.3, 4.05, 0.54, 7, 0.16, 0.76, 3.36, 2.7, 8.34, 4.09, 6.89, 31.68, 19.55, 3.16, 15.02, 13.1, 1.33, 1.06, 11.34, 4.94, 1.84, 4, 1.38, 7.7, 3.69, 1.29, 11.66, 1.72, 2.59, 13.45, 5.98, 6.9, 6.05, 4.96, 0.94, 24.56, 6.21, 2.33, 3.23, 0.99, 4.55, 20.86, 3.07, 18.07, 25.94, 8.51, 0.52, 2.27, 3.58, 4.31, 3.56, 3.04, 3.56, 1.21, 2.09, 3.63, 2.56, 7.8, 7.06, 13.48, 24.77, 2.3, 74.65, 19.44, 3.35, 13.96, 8, 5.22, 1.23, 19.89, 18.59, 18.59, 16.25, 15.28, 9.18, 9.1, 13.5, 13.99, 6.39, 10.39, 1.18, 10.24, 4.06, 21.32, 30.84, 9.17, 14.8, 24.77, 50.12, 50.92, 5.37, 17.77, 22.03, 24.07, 11.94, 3.67, 8.07, 20.49, 19.79, 6.37, 7.43, 7.4, 53.35, 7.77, 10.6, 10.6, 7.42, 4.83, 8.03, 12.4, 14, 4.64, 13.05, 17.47, 5.67, 23.77, 29.97, 17.78, 21.05, 33.57, 2.14, 18.43, 41.84, 10.59, 0.5, 18.8, 11.45, 16.54, 25.89, 58.63, 21.31, 0.49, 23.86, 36.37, 41.89, 0.5, 38.84, 52.92, 27, 12.84, 53.4, 16.58, 16.14, 22.01, 10.04, 19.94, 19.06, 8.03, 12.48, 1.84, 3, 3.16, 31.87, 1.95, 35.95, 20.57, 8.13, 27.91, 4.78, 9.43, 5.11, 1.85, 29.39, 15.79, 8.17, 10.57, 1.2, 1.22, 10.95, 12.15, 3.62, 2.73, 5.21, 3.41, 0.63, 4.67, 7.23, 3.43, 1.77, 5.25, 4.12, 4.61, 1.33, 10.91, 10.15, 11.51, 26.77, 17.62, 22.53, 28.32, 4.36, 29.96, 12.31, 11.56, 11.56, 11.56, 11.56, 11.56, 11.56, 11.56)
out_BM<-mixmodCluster(BM, nbCluster=2:8)
out_BM
hist(out_BM)
le output et le graphique obtenu :
Code : Tout sélectionner
****************************************
*** INPUT:
****************************************
* nbCluster = 2 3 4 5 6 7 8
* criterion = BIC
****************************************
*** MIXMOD Models:
* list = Gaussian_pk_Lk_C
* This list includes only models with free proportions.
****************************************
* data (limited to a 10x10 matrix) =
[1] 9.19 0.5 10.61 7.98 10.34 6.03 14.6 39.82 8.07 9.33
* ... ...
****************************************
*** MIXMOD Strategy:
* algorithm = EM
* number of tries = 1
* number of iterations = 200
* epsilon = 0.001
*** Initialization strategy:
* algorithm = smallEM
* number of tries = 10
* number of iterations = 5
* epsilon = 0.001
* seed = NULL
****************************************
****************************************
*** BEST MODEL OUTPUT:
*** According to the BIC criterion
****************************************
* nbCluster = 4
* model name = Gaussian_pk_Lk_C
* criterion = BIC(2812.4128)
* likelihood = -1373.2533
****************************************
*** Cluster 1
* proportion = 0.3948
* means = 8.4066
* variances = 8.9541
*** Cluster 2
* proportion = 0.0643
* means = 44.2755
* variances = 378.7400
*** Cluster 3
* proportion = 0.2839
* means = 2.7419
* variances = 2.2573
*** Cluster 4
* proportion = 0.2571
* means = 17.9805
* variances = 45.0015
****************************************
Et voici ma question :
En lisant l'output de mixmodCluster j'ai remarqué qu'il y avait de noté "data (limited to a 10x10 matrix)" or mes valeurs entrent dans une matrice en 20x20.
Mon modèle est donc tronqué ? Je fais comment pour passer en 20x20 ?