Before plunging in make sure you know whether you are referencing data or statistics. Unfortunately, the terms are sometimes used as though they are interchangeable.
This is not the case: "statistics are the interpretation and summary of data" while "data are raw information from which statistics are derived" ("Data or Statistics?" Finding and Using Data for your Research).
The most important thing to remember is that you want your citation to include enough information so that a reader could find the same dataset again in the future, even if the link you provide no longer works. It's necessary to include a mixture of general and specific information to help them be certain that they've found the same dataset that you were referring to.
Citing data has not always been standard practice, especially if it is data you have collected yourself, but as data becomes more and more widely shared proper attribution is increasingly important. Citing datasets helps them become part of the scholarly record and gives proper credit to the creator of the dataset.