Mark Williamson wrote:
A good, comprehensive study is needed (and probably already exists) to better classify the varieties of Arabic, because SIL has done a piss-poor job. It's shameful, considering just how many people speak one or another variety of Arabic.
It does exist, but the problem is that dozens of such studies exist, arriving at different results. =] There is also disagreement over the extent to which it even makes sense to classify Arabic into distinct varieties, given the complex varieties of diglossia (to the point of "multiglossia"), continuums along multiple axes of variation, and discrepancies between speakers' reported dialect/language and actual grammatical features.
A decent survey can be found in the editor's introduction to the collection _Understanding Arabic_ (1996, ed. A. Elgibali, ISBN 9774243722), and in more details in some of the papers therein.
Of course, I'm no expert in the field; have just read a small amount of what's written by people who are. It's a complex enough area that it's basically an entire field of research.
-Mark