Search
Question: Error using function NbClust regarding multiple values in the argument "index" of package NbClust for getting the optimal number of clusters of a dataset
1
gravatar for svlachavas
2.2 years ago by
svlachavas560
Greece/Athens/National Hellenic Research Foundation
svlachavas560 wrote:

Dear Bioconductor community,

i would like to adress a specific problem i encountered after using the package NbClust, for getting the optimal number of clusters concerning various measures, regarding a microarray dataset. Although the problem is possibly more generally programming-dependent, i decide to create the post as it is directly connected with the specific package.

In detail, the function Nblust according to the documents is used:

nc <- NbClust(exprs(eset.3), min.nc=2, max.nc=15, method="kmeans", index="all") # eset.3 my expression set & also the argument index contains all of the different criteria used for evaluation of the optimal number of clusters for each methodology:

Error in NbClust(exprs(eset.3), min.nc = 2, max.nc = 15, method = "kmeans") : 

The TSS matrix is indefinite. There must be too many missing values. The index cannot be calculated.

While i solved this problem from literature and also from feedback from the creators of the package, as it is probably due to the fact that my matrix returns negative eigenvalues--Nevertheless, except some specific criteria i cant use for my set and which create the above error, i can use the rest 23 specific indexes with the function,

i.e.[ "kl", "ch", "hartigan",  "cindex"...]

The new (unsolved) problem is when i tried to directly used it in the function with only a couple of indices:

 nc <- NbClust(df, min.nc=2, max.nc=15, method="kmeans", index=c("kl","ch","hartigan",  "cindex", "db"))


Error in NbClust(df, min.nc = 2, max.nc = 15, method = "kmeans", index = c("kl",  : 
  object 'DiffLev' not found
In addition: Warning messages:
1: In if is.na(indice)) stop("invalid clustering index") :
  the condition has length > 1 and only the first element will be used
2: In if (indice == -1) stop("ambiguous index") :
  the condition has length > 1 and only the first element will be used

So i assume that because the values for the index argument are fixed, and either use one index value at a time(i.e. "ch") or the options "all" & "alllong"(it shows on ?NbClust), i tried to create a function that uses each value from the pre-defined values with a for loop and returns a list which contains each list for the specific index, based on the fact that :

the function NbClust returns a list, with 4 components, from which the one i want is called $Best.nc, returns the optimum number of clusters according to each criterion-and to conclude, i would like to create generally an object that for each criterion-index from the above will return also this specific value .

Thus, i naively tried:

selected <- c( "kl", "ch", "hartigan",  "cindex", "db", "silhouette", "duda", "pseudot2", "beale", "ratkowsky", "ball", "ptbiserial", "gap", "frey", "mcclain", "gamma", "gplus", "tau", "dunn", "hubert", "sdindex", "dindex", "sdbw") # the specific indices i can use-total number=23

results <- vector("list",23)

for (i in 1:length(selected)) {results[[i]] <- NbClust(exprs(eset.2), min.nc=2, max.nc=15, method="kmeans", index=selected[i]) }

But the loop keeps running without stopping like possibly an infinite loop(i left it a couple of hours but nothing happen)

Possibly, it is due to my  lack of experienced programming, so how could i deal with this problem ?

Any help or suggestions would be essential !!

 

 

 

 

 

ADD COMMENTlink modified 7 months ago by james.tsakalos0 • written 2.2 years ago by svlachavas560
0
gravatar for james.tsakalos
7 months ago by
james.tsakalos0 wrote:

I know this is an old post, I just had same issue.

 

My solution was to wrap try() around the loop then to look at the results.

 

for (i in 1:length(selected)) {

results[[i]] <- try(NbClust(exprs(eset.2), min.nc=2, max.nc=15, method="kmeans", index=selected[i]))

}

 

This should work for you now.

ADD COMMENTlink written 7 months ago by james.tsakalos0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 149 users visited in the last hour