I had a list of 1100 words and sought to cluster them based on a levenshtein distance of 2. This reduced the list to 900 words. Running the calculation again reduced the number to roughly 800, and again reduced it to about 750. I was surprised that the change was not more dramatic
However, if I cluster the 1100 words based on a levenshtein distance of 6 from the start, they are reduced to about 400.
Seemingly the two methods are not at all interchangeable. Unless there is something wrong with my program, what makes the iterative Levenshtein distance calculation so different from the single instance, and what implications might this bear on the iterative process?