2

I wish to use the permutation distribution clustering package of R (https://cran.r-project.org/web/packages/pdc/pdc.pdf) for multivariate time series clustering. After using pdclust method(page 11 of the url pdf ) for hierarchical clustering, I plotted the dendogram using plot method (page 11 again). There are 60 samples. So, in the plot (Hierarchical clustering plot ), there are 60 time series but they are unlabelled. When I try to specify a vector of labels instead of labels = NULL, I always get this error "Error in graphics:::plotHclust(n1, merge, height, order(x$order), hang, : invalid dendrogram input". Any help will be appreciated. Here is my code:

data1  <- read.csv(file="file_PID_1_1Apr_00-03.csv",head=FALSE,sep=",")
data2  <- read.csv(file="file_PID_2_1Apr_00-03.csv",head=FALSE,sep=",")
data3  <- read.csv(file="file_PID_3_1Apr_00-03.csv",head=FALSE,sep=",")
data4  <- read.csv(file="file_PID_4_1Apr_00-03.csv",head=FALSE,sep=",")
data5  <- read.csv(file="file_PID_5_1Apr_00-03.csv",head=FALSE,sep=",")
data6  <- read.csv(file="file_PID_6_1Apr_00-03.csv",head=FALSE,sep=",")
data7  <- read.csv(file="file_PID_7_1Apr_00-03.csv",head=FALSE,sep=",")
data8  <- read.csv(file="file_PID_8_1Apr_00-03.csv",head=FALSE,sep=",")

data9  <- read.csv(file="file_PID_1_1Apr_03-06.csv",head=FALSE,sep=",")
data10 <- read.csv(file="file_PID_2_1Apr_03-06.csv",head=FALSE,sep=",")
data11 <- read.csv(file="file_PID_3_1Apr_03-06.csv",head=FALSE,sep=",")
data12 <- read.csv(file="file_PID_4_1Apr_03-06.csv",head=FALSE,sep=",")
data13 <- read.csv(file="file_PID_5_1Apr_03-06.csv",head=FALSE,sep=",")
data14 <- read.csv(file="file_PID_6_1Apr_03-06.csv",head=FALSE,sep=",")
data15 <- read.csv(file="file_PID_7_1Apr_03-06.csv",head=FALSE,sep=",")
data16 <- read.csv(file="file_PID_8_1Apr_03-06.csv",head=FALSE,sep=",")

data17 <- read.csv(file="file_PID_1_1Apr_06-09.csv",head=FALSE,sep=",")
data18 <- read.csv(file="file_PID_2_1Apr_06-09.csv",head=FALSE,sep=",")
data19 <- read.csv(file="file_PID_3_1Apr_06-09.csv",head=FALSE,sep=",")
data20 <- read.csv(file="file_PID_4_1Apr_06-09.csv",head=FALSE,sep=",")
data21 <- read.csv(file="file_PID_5_1Apr_06-09.csv",head=FALSE,sep=",")
data22 <- read.csv(file="file_PID_6_1Apr_06-09.csv",head=FALSE,sep=",")
data23 <- read.csv(file="file_PID_7_1Apr_06-09.csv",head=FALSE,sep=",")
data24 <- read.csv(file="file_PID_8_1Apr_06-09.csv",head=FALSE,sep=",")

data25 <- read.csv(file="file_PID_1_1Apr_09-12.csv",head=FALSE,sep=",")
data26 <- read.csv(file="file_PID_2_1Apr_09-12.csv",head=FALSE,sep=",")
data27 <- read.csv(file="file_PID_3_1Apr_09-12.csv",head=FALSE,sep=",")
data28 <- read.csv(file="file_PID_4_1Apr_09-12.csv",head=FALSE,sep=",")
data29 <- read.csv(file="file_PID_5_1Apr_09-12.csv",head=FALSE,sep=",")
data30 <- read.csv(file="file_PID_6_1Apr_09-12.csv",head=FALSE,sep=",")
data31 <- read.csv(file="file_PID_7_1Apr_09-12.csv",head=FALSE,sep=",")
data32 <- read.csv(file="file_PID_8_1Apr_09-12.csv",head=FALSE,sep=",")

data33 <- read.csv(file="file_PID_1_1Apr_12-15.csv",head=FALSE,sep=",")
data34 <- read.csv(file="file_PID_2_1Apr_12-15.csv",head=FALSE,sep=",")
data35 <- read.csv(file="file_PID_3_1Apr_12-15.csv",head=FALSE,sep=",")
data36 <- read.csv(file="file_PID_4_1Apr_12-15.csv",head=FALSE,sep=",")
data37 <- read.csv(file="file_PID_5_1Apr_12-15.csv",head=FALSE,sep=",")
data38 <- read.csv(file="file_PID_6_1Apr_12-15.csv",head=FALSE,sep=",")
data39 <- read.csv(file="file_PID_7_1Apr_12-15.csv",head=FALSE,sep=",")
data40 <- read.csv(file="file_PID_8_1Apr_12-15.csv",head=FALSE,sep=",")

data41 <- read.csv(file="file_PID_2_1Apr_15-18.csv",head=FALSE,sep=",")
data42 <- read.csv(file="file_PID_3_1Apr_15-18.csv",head=FALSE,sep=",")
data43 <- read.csv(file="file_PID_4_1Apr_15-18.csv",head=FALSE,sep=",")
data44 <- read.csv(file="file_PID_6_1Apr_15-18.csv",head=FALSE,sep=",")
data45 <- read.csv(file="file_PID_7_1Apr_15-18.csv",head=FALSE,sep=",")
data46 <- read.csv(file="file_PID_8_1Apr_15-18.csv",head=FALSE,sep=",")

data47 <- read.csv(file="file_PID_1_1Apr_18-21.csv",head=FALSE,sep=",")
data48 <- read.csv(file="file_PID_2_1Apr_18-21.csv",head=FALSE,sep=",")
data49 <- read.csv(file="file_PID_3_1Apr_18-21.csv",head=FALSE,sep=",")
data50 <- read.csv(file="file_PID_4_1Apr_18-21.csv",head=FALSE,sep=",")
data51 <- read.csv(file="file_PID_6_1Apr_18-21.csv",head=FALSE,sep=",")
data52 <- read.csv(file="file_PID_7_1Apr_18-21.csv",head=FALSE,sep=",")
data53 <- read.csv(file="file_PID_8_1Apr_18-21.csv",head=FALSE,sep=",")

data54 <- read.csv(file="file_PID_1_1Apr_21-24.csv",head=FALSE,sep=",")
data55 <- read.csv(file="file_PID_2_1Apr_21-24.csv",head=FALSE,sep=",")
data56 <- read.csv(file="file_PID_3_1Apr_21-24.csv",head=FALSE,sep=",")
data57 <- read.csv(file="file_PID_4_1Apr_21-24.csv",head=FALSE,sep=",")
data58 <- read.csv(file="file_PID_6_1Apr_21-24.csv",head=FALSE,sep=",")
data59 <- read.csv(file="file_PID_7_1Apr_21-24.csv",head=FALSE,sep=",")
data60 <- read.csv(file="file_PID_8_1Apr_21-24.csv",head=FALSE,sep=",")





list <- array(0,dim=c(720,60,4))

myfunc <- function(j,i,k){
    if (j == 1) return (data1[i,k]) 
    else if (j==2) return (data2[i,k])
    else if (j==3) return (data17[i,k])
    else if (j==4) return (data9[i,k])
    else if (j==5) return (data5[i,k])
    else if (j==6) return (data6[i,k])
    else if (j==7) return (data7[i,k])
    else if (j==8) return (data8[i,k])
    else if (j==9) return (data9[i,k])
    else if (j==10) return (data10[i,k])
    else if (j==11) return (data11[i,k])
    else if (j==12) return (data12[i,k])
    else if (j==13) return (data13[i,k])
    else if (j==14) return (data14[i,k])
    else if (j==15) return (data15[i,k])
    else if (j==16) return (data16[i,k])
    else if (j==17) return (data17[i,k])
    else if (j==18) return (data18[i,k])
    else if (j==19) return (data19[i,k])
    else if (j==20) return (data20[i,k])
    else if (j==21) return (data21[i,k])
    else if (j==22) return (data22[i,k])
    else if (j==23) return (data23[i,k])
    else if (j==24) return (data24[i,k])
    else if (j==25) return (data25[i,k])
    else if (j==26) return (data26[i,k])
    else if (j==27) return (data27[i,k])
    else if (j==28) return (data28[i,k])
    else if (j==29) return (data29[i,k])
    else if (j==30) return (data30[i,k])
    else if (j==31) return (data31[i,k])
    else if (j==32) return (data32[i,k])
    else if (j==33) return (data33[i,k])
    else if (j==34) return (data34[i,k])
    else if (j==35) return (data35[i,k])
    else if (j==36) return (data36[i,k])
    else if (j==37) return (data37[i,k])
    else if (j==38) return (data38[i,k])
    else if (j==39) return (data39[i,k])
    else if (j==40) return (data40[i,k])
    else if (j==41) return (data41[i,k])
    else if (j==42) return (data42[i,k])
    else if (j==43) return (data43[i,k])
    else if (j==44) return (data44[i,k])
    else if (j==45) return (data45[i,k])
    else if (j==46) return (data46[i,k])
    else if (j==47) return (data47[i,k])
    else if (j==48) return (data48[i,k])
    else if (j==49) return (data49[i,k])
    else if (j==50) return (data50[i,k])
    else if (j==51) return (data51[i,k])
    else if (j==52) return (data52[i,k])
    else if (j==53) return (data53[i,k])
    else if (j==54) return (data54[i,k])
    else if (j==55) return (data55[i,k])
    else if (j==56) return (data56[i,k])
    else if (j==57) return (data57[i,k])
    else if (j==58) return (data58[i,k])
    else if (j==59) return (data59[i,k])
    else if (j==60) return (data60[i,k])

}

list <- array(0,dim=c(720,60,4))


for(i in 1:720){
    for (j in 1:60){
        list[i,j,1] <- myfunc(j,i,6)
        list[i,j,2] <- myfunc(j,i,7)
        list[i,j,3] <- myfunc(j,i,8)
        list[i,j,4] <- myfunc(j,i,9)
    }
}

library("pdc")
clustering <- pdclust(list)
plot(clustering, labels= NULL, type="rectangle", timeseries.as.labels = T, p.values=T)
  • Sometimes it helps when you already give some headlines to the input data, before clustering and before plotting, but only a guess ... – holzben Jul 17 '16 at 10:56
  • 1
    I tried it but it didn't help –  Jul 17 '16 at 11:46

1 Answers1

2

I created a slim version of your code that works without the data files, so that it is easier to talk about your problem. Here, I create 60 time series with 4 dimensions and 720 time points (like you do). Only that I simulate half of trials from a random normal and the other half from a linear trend with a superimposed random normal. So, they should clearly be separable into two groups for pdc. Here is the code:

require("pdc")

# make this replicable by setting a random seed
set.seed(7823)

# 60 TS each with 4 dimensions and 720 timepoints
# half of them are random uniform other half are a mix of random uniform 
# and linear increase
list <- array(0,dim=c(720,60,4))
for (i in 1:30) {
  for (j in 1:4) {
    list[,i,j] <- rnorm(n = 720)
    list[,i+30,j] <- rnorm(n=720)+1:720
  }
}
cols <- c(rep("red",30),rep("blue",30))
labels <- c(rep("normal",30),rep("normal+trend",30))

# run clustering and color original groups each in red and blue
clustering <- pdclust(list)

pdf("pdcplot.pdf")
plot(clustering, labels= labels, type="rectangle", cols=cols, cex=0.5)
dev.off()

I have no troubles plotting the labels. I added a "cex=0.5" to decrease font size in the plot. Also, I removed the "timeseries.as.labels = T" since this is overridden when you specify labels. Here is how my plot looks like (with labels):

Clustering of simulated data with labels

I can only reproduce the error you reported when the number of labels I specify does not match the number of time series. You may want to check the size of your label vector again (e.g., length(labels)==60).

Brandmaier
  • 134
  • 4