2

I've only been using R for a short time. I use R 3.4.4. For a study I created clusters as follows:

library('cluster')
clusterward2 <- agnes(dist.om2, diss = TRUE, method = "ward")
plot(clusterward2)
plot(sort(clusterward2$height, decreasing=TRUE)[1:15], type='s', xlab="nb de classes", ylab="inertie")
points(c(2, 3,5), sort(clusterward2$height, decreasing=TRUE)[c(2, 3,5)],
       col = c("green3", "red3", "blue3"), cex = 2, lwd = 4)
cl2.3 <- cutree(clusterward2, k = 3)

Then, i retrieve the individual belonging to each cluster as follow :

split(mydata$colonneID, cl2.3)

Is it possible to retrieve the score that each individual had for clustering ? I would like to analyse the extreme of each class, but I don't know if it is possible and how do it.

My data : 

donnees <- "CT  H_NH    I1  I2  I3  I4
CT_5    humain  SN_def  SN_dem  SN_dem  Pro
CT_6    humain  SN_def  SN_ind  SN_def  SN_dem
CT_7    humain  SN_def  SN_dem  SN_pos  SN_dem
CT_8    humain  SN_def  Autre   SN_def  SN_def
CT_9    humain  Autre   SN_def  SN_def  SN_def
CT_15   humain  SN_ind  SN_def  SN_def  SN_def
CT_17   humain  Autre   SN_pos  SN_dem  Autre
CT_18   humain  SN_def  Pro SN_def  Pro
CT_19   humain  SN_def  Pro SN_def  SN_pos
CT_20   humain  SN_def  SN_def  Pro SN_pos
CT_27   humain  NPP Pro Pro Pro
CT_29   humain  NPP SN_sansDET  NPP SN_pos
CT_30   humain  SN_sansDET  Pro SN_def  Pro
CT_32   humain  SN_def  SN_def  SN_def  SN_dem
CT_33   humain  Autre   NPP NPP SN_def
CT_34   humain  NPP Pro NPP Pro
CT_35   humain  SN_def  NPP Pro NPP"

Data <- read.table(text=donnees, header = TRUE)

The code i used : 

Data.lab <- seqstatl(Data[,3:6])

Data.scode <- c("Autre", "NPP", "Pro", "SN_def", "SN_dem", "SN_ind", "SN_pos", "SN_sansDET")

Data.seq_7 <- seqdef(Data[, 3:6], states = Data.scode)

submat2 <- seqsubm(Data.seq_7, method = "TRATE")

dist.om2 <- seqdist(Data.seq_7, method = "OM", indel = 1,  sm = submat2)
library('cluster')
clusterward2 <- agnes(dist.om2, diss = TRUE, method = "ward")
plot(clusterward2, which.plots=2)
plot(sort(clusterward2$height, decreasing=TRUE)[1:15], type='s', xlab="nb de  classes", ylab="inertie")
points(c(2, 3,5), sort(clusterward2$height, decreasing=TRUE)[c(2, 3,5)],
 col = c("green3", "red3", "blue3"), cex = 2, lwd = 4)
cl2.3 <- cutree(clusterward2, k = 3)

Thank you very much for your help.

Gilbert
  • 3,570
  • 18
  • 28
Silvia Federzoni
  • 21
  • 3

1 Answers1

1

You get the distances to the cluster centers with the disscenter function of TraMineR

diss.to.cl.center <- disscenter(dist.om2, group=cl2.3)
res <- data.frame(seqconc(Data.seq_7), cl2.3, diss.to.cl.center)
res

#                       Sequence cl2.3 diss.to.cl.center
# 1     SN_def-SN_dem-SN_dem-Pro     1         0.9947090
# 2  SN_def-SN_ind-SN_def-SN_dem     2         1.3511905
# 3  SN_def-SN_dem-SN_pos-SN_dem     1         0.7447090
# 4   SN_def-Autre-SN_def-SN_def     2         1.3167989
# 5   Autre-SN_def-SN_def-SN_def     2         1.3167989
# 6  SN_ind-SN_def-SN_def-SN_def     2         1.3167989
# 7    Autre-SN_pos-SN_dem-Autre     1         1.4788360
# 8        SN_def-Pro-SN_def-Pro     2         1.2096561
# 9     SN_def-Pro-SN_def-SN_pos     2         1.1957672
# 10    SN_def-SN_def-Pro-SN_pos     2         1.1554233
# 11             NPP-Pro-Pro-Pro     3         1.4152381
# 12   NPP-SN_sansDET-NPP-SN_pos     3         1.7533333
# 13   SN_sansDET-Pro-SN_def-Pro     2         2.3842593
# 14 SN_def-SN_def-SN_def-SN_dem     2         0.7890212
# 15        Autre-NPP-NPP-SN_def     3         1.7485714
# 16             NPP-Pro-NPP-Pro     3         0.6652381
# 17          SN_def-NPP-Pro-NPP     3         1.1033333
Gilbert
  • 3,570
  • 18
  • 28