Comparing subsets of a population with a very different size (power law case)?


New Member
My population (12000 data) is characterized by events that are either of type A or of type B. My goal is to figure out whether the relative proportion of A with respect to B does change if a property of the events (the size in my case) change.
My first approach was quite simple: dividing the data in different classes, each corresponding to a specific size range, and then calculating the proportion for each class.
The problem is that the property of the events under consideration is distributed according to a power law with respect to the frequency of the events: in few words, there are a lot of large events and very few small ones. This yields an issue about the statistical significance of my first approach. In fact, the proportion of events of type A is calculated on the basis of very few data for the last class (which includes the largest events). This means I am mistakenly drawing conclusions about how the ratio A/B varies according to the size of the events.
Do you have any suggestions about how to approach this issue?
Thank you!