0

I am using clusplot to plot PCs using kmeans, but the tricky part is, i want to have the ellipses color coded by a different variable (NOT the f$cluster, which is used for data points). Here is the command that i am giving, but something is not working right here!

    clusplot(f[,1:3],f$cluster,cex=0.8,color=TRUE,shade=TRUE, 
     labels=5,
     col.clus=unique(f$groupnum),
     lines=0,main="testing")

my f is a dataframe with 33 columns. PC1 to PC30, followed by cluster, groupnum and groupttl. $cluster codes for clustering, but i am trying to use $groupnum for color coding and possible $groupttl for labeling the ellipses.

One important thing is: Both $cluster and $groupnum can have different #groups (i.e) i have only 5 clusters but 6 unique groupnum values to be colored for ellipses. Therefore, it is usually the case, that ellipses to be colored will be more than actual number of clusters.

Thank you Paul and shuja. Here is my data frame


structure(list(PC1 = c(8.2545, 1.0159, -1.5319, -1.7703, 1.6903, 
-1.7378, -2.1898, -2.8501, 1.0669, -0.8577, 6.3985, 2.6495, 2.6374, 
0.2395, 8.6415, -0.0638, -2.739, -0.7897, 0.3137, -3.0958, 2.9457, 
-4.1256, -2.6289, -2.9377, -0.6272, -2.7296, 1.2871, -1.6917, 
-0.1118, 11.1845, 2.6486, -3.4377, -0.5581, -3.217, -2.5425, 
-1.419, -0.9338, -0.8993, -3.119, -3.5188, 1.6804, -0.4142, -2.3187, 
-0.1962, -1.3428, 5.1539, 10.3632, -0.9815, -2.7796, 0.0708), 
    PC2 = c(-1.6021, -5.4785, 3.3933, 3.6795, -3.0405, 5.3614, 
    -2.895, 2.2553, -2.648, -0.451, -1.3402, -2.6696, -4.3236, 
    0.0659, -0.4115, 4.0168, 0.034, -0.2273, -0.5867, 0.8339, 
    -0.2328, 1.5119, 5.183, -0.7078, 0.5813, 2.8371, 0.8223, 
    -1.2817, 0.0378, -3.297, -1.0233, 0.5048, -1.9093, -5.5851, 
    -0.8716, -2.135, -4.2768, 1.567, -0.1263, 2.4107, -1.3151, 
    1.6173, -0.3908, 3.7365, -1.3812, -2.1328, -0.05, 0.849, 
    1.9369, -4.7095), PC3 = c(1.4114, -2.0719, -1.927, -0.417, 
    -2.2733, -1.8481, -2.807, 0.8132, -2.6583, -0.5894, 1.4173, 
    2.9953, 0.2831, -1.0971, 2.7126, -3.8635, -0.5739, 2.5493, 
    1.3207, 2.3459, 4.5259, 1.6239, -0.1763, 1.929, 1.5237, 1.7709, 
    2.1231, -1.5679, 2.9978, -1.0623, 6.3311, 3.2371, -0.0466, 
    -3.2293, 0.3979, -3.4121, -1.6269, 0.8722, -0.2534, 2.3849, 
    -1.6068, -0.8486, 0.9351, -1.8844, 2.8963, 0.4948, 1.3549, 
    0.522, 2.3628, -2.2726), cluster = c(5L, 2L, 1L, 1L, 2L, 
    1L, 2L, 4L, 2L, 3L, 5L, 3L, 2L, 3L, 5L, 1L, 4L, 3L, 3L, 4L, 
    3L, 4L, 1L, 4L, 3L, 4L, 3L, 2L, 3L, 5L, 3L, 4L, 2L, 2L, 4L, 
    2L, 2L, 3L, 4L, 4L, 2L, 1L, 4L, 1L, 3L, 5L, 5L, 3L, 4L, 2L
    ), groupnum = c(8L, 9L, 2L, 2L, 4L, 2L, 4L, 10L, 4L, 6L, 
    8L, 8L, 9L, 2L, 8L, 2L, 9L, 6L, 6L, 10L, 8L, 10L, 10L, 10L, 
    2L, 10L, 6L, 4L, 6L, 8L, 8L, 6L, 9L, 9L, 10L, 4L, 9L, 6L, 
    9L, 10L, 4L, 2L, 9L, 2L, 2L, 4L, 8L, 6L, 6L, 9L), groupttl = c("MN", 
    "GC", "GMS", "GMS", "MP", "GMS", "MP", "MS", "MP", "BC", 
    "MN", "MN", "GC", "GMS", "MN", "GMS", "GC", "BC", "BC", "MS", 
    "MN", "MS", "MS", "MS", "GMS", "MS", "BC", "MP", "BC", "MN", 
    "MN", "BC", "GC", "GC", "MS", "MP", "GC", "BC", "GC", "MS", 
    "MP", "GMS", "GC", "GMS", "GMS", "MP", "MN", "BC", "BC", 
    "GC")), .Names = c("PC1", "PC2", "PC3", "cluster", "groupnum", 
"groupttl"), row.names = c(NA, 50L), class = "data.frame")

I look forward for some help on this!

Thanks SP

user2105887
  • 5
  • 1
  • 5
  • 1
    See http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example for making a reproducible example. If your data.frame is very large, I suggest you just post a subset, enough needed to reproduce your problem. – Paul Hiemstra Apr 01 '13 at 16:16
  • I am not able to post even 10 rows of my data frame using 
    . it is constantly asking me to cut down the #characters. Is there a way to attach my file. there is a learning curve for posting stuff here, like for every thing. If you can point me to one best way, that i can post atleast 50 lines of my data, i will stick with that!
     – user2105887 Apr 01 '13 at 19:58
  • @user2105887 You could round your data `ff <- f`, then `ff[, 2:4] <- round(ff[, 2:4], 4)`, then you could post `dput(head(ff, 20))`. Using `dput` is nice (as in Paul's link) because someone else can copy/paste it into their R session. – Gregor Thomas Apr 01 '13 at 22:23
  • And posting just a few columns, as you did, is good. – Gregor Thomas Apr 01 '13 at 22:23

0 Answers0