With Respect
The index of dispersion, variance divided by mean, is normally used to check for, among other things, clustering, in counts of something within equal sized areas of space or units of time.
I tried calculating it for a set of 46 cases, each with a different count of its citations by other cases, looking for signs that cases clustered at particular numbers of citations. I also calculated it for six subsets of cases.
The overall index is 163.3802, indicating very great overdispersion and the posibility of extensive clustering. The subset indexes range from 0.424125 to 13.18216, three being much under 1 and three much over it. One of the three shown as overdispersed is a group of two cases, which hardly can have clusters.
Does this measure not work on data for citations per case versus units of space and time. If so, why?
Though my subsets were of different numbers of cases, I was measuring dispersion over units of the same size, one case, in all calculations.
Yours Sincerely,
Alan E. Dunne
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|