Hi Jason,
If your faxes are time stamped, then I think you can take a sample by date. Depending on the length of time you collected these faxes ( for example, over a month, year, week, few days, etc.). You could use random systemic sampling (e.g. sample every third day) and over time you should have a representative sample. If you know of certain behaviours, such as more faxes on Mondays than on Fridays, that could help you decide on your sampling pattern.
It is suggested to do two sets of samples in this manner, apply your tests on each group – if they are vastly different, you should re-think your samples. But if tests show the groups are similar, then you have a safer set of results that are based upon the true population.
Hope that helps!
.
