Interpretation of low, negative, cubic clustering criterion values

Hello all,

My question is how do you use the cubic clustering criterion to assist in selecting an appropriate number of clusters when the values are all negative?

I know that when they are positive you look for peaks in the values to indicate a demarcation point. However, when they are all negative values do you look then for troughs/low points or still peaks? These negative values are only low (max -3) not excessively high indicating outliers.

I can't seem to find any information on this matter other than on the fact the distribution may be unimodal or long tailed.

Your advice is much appreciated,