Estimate the average number of visitors per page(websites) from a page view data stream of Tuples when 20% accessibility of the entire stream

Let us assume that we have access to page view data from several websites in the form of a stream of tuples with the following schema: (Domain_ID, Page_ID, Person_ID)

Let's make the following assumptions about this stream:

Domain_IDs are unique, but a Page_ID is unique only within a Domain (i.e., different Domains may assign the same ID to different pages)

Person ID's are unique only within a Domain (different Domains may assign the same ID to different visitors).

In the full input stream, for each Domain we have S # of pages visited by N visitors each and D # of pages visited by 2N visitors each and no pages have a different numbers of visitors.

There are no duplicate visits by any user for any page, i.e. a given (Domain_ID, Page_ID, Person_ID) tuple only occurs once in the entire stream

We are interested in answering the following query: - For each domain, estimate the average number of visitors per page.

If we only have access to a random sample that captures 20% of the entire input stream, estimate the degree of underestimation or overestimation that this represents as compared to the true averages in the stream.