Dice's similarity coefficient (aka nei and li). significance...

I've been struggling with the Dice coefficient for days now. I understand that the closer to 1 the cases are the more similar they are but is there some way of obtaining a p value for this to show some level of how significant this similarity is to one another? i can't seem to find anything about it atall in books, papers or internet sites and so i fall upon your mercy!

i have 40 individuals from a population. for each individual i have presence/absence data for 47 potential 'peaks' from RAPD analysis (using 4 different short primers). so each individual essentially ends up with a binary code of 47 digits. i want to know for all 40 individuals how similar they are to each other (e.g. how similar are individual 1 and 2, individual 1 and 3, individual 1 and 4 etc to individual 39 and 40) and the probability that they are significantly similar / significantly different from each other.

reason for using dice: reading through a fair few papers it seems that Dice's coefficient is the most suitable for data gathered in this manner, reducing potential errors incurred from the RAPD runs.

any advice atall or information on where else to look for this information would be really helpful! i've been at this for days now and am starting to worry that either a. there is no way to calculate the probability that they are signficantly similar/different or b. i've got too involved in this hunt and am missing something really obvious!

Many thanks
i have just encountered this problem these days. i consulted many people, but they were either unknow about the Dice coefficient or didn't know how to collect these data.

have you solved this problem yet?