You can work it all out with this calculator. Tip: You can put different types of MTT's in a single calculation to see how best to balance higher variance types with lower variance and work out what you are comfortable with.

http://www.runitonce.com/pro-training/videos/bankroll-management-and-variance/

Conceptually the term of variance does not have much to say about the larger the field. Usually, the larger the field more players are bad and you can decrease your variance by selecting better the spots to play against them. On smaller fields, which gives you a better shot to take the final table but usually at 100+ the fields get a lot tougher, which increases your variance. 

@Raphael: Variance does depend on field size. You make a good point that the large field MTT's have worse players and therefore our ROI should be higher. At a certain improved ROI our variance will decrease because the increased ROI overcomes the variance caused by the larger field.

The factors involved in differences in variance are: ROI, field size and prize structure.

You can think about edge cases like winner take all or only final table paid MTT's and how difficult it would be to realise even huge ROI's reliably, in big fields with these structures and compare that to the very small ROI's the best HUSNG hyper players make a great living from.

In any event, the tournament variance calculator (link in earlier post) works it all out for you - the only hard part is estimating your ROI accurately in a given format.

Variance is how far, in average, you are from your ROI. I can not disagree but I can not agree 100%. 

I like this question and have thought long and hard about this myself. In practice i have tried changing entire sessions from a lot of $2-$27 larger fields, and replaced them with the likes of say 20 or so of the smaller fields you mention eg $33 1r1a and 55fo etc..

what i did find was this..

Those fields were 3/4 full of good regs and either a) i was just finding it way too difficult to accumulate chips from or b) i wasnt adjusting my game well enough to accomodate for the field change.

I accept that i have done this for probably <10 total sessions (because the downswing in the $55-$109 fields was ridic for my roll) so the sample size was hardly significant but during those games i did notice quite a difference in the calibre of the 'average' player there.

Just my observations for ya:)

Thanks for the replies all.

The variance calculator above was very useful and sort of confirmed my suspicion that over the medium term, variance in smaller fields with bigger buyins is actually less than if you were playing stuff like the daily bigs etc with fields in the low-to-mid-thousands. This was factoring in a significantly lower ROI in the midstakes tourneys too due to the overall tougher fields. 

@thephyzician - the point about fields is a valid one, although when I've played these games myself, or railed friends in them, there are still some pretty bad players in them. Luck of the seat draw I guess. The last 109 freezeout I played, all 8 opponents were big winning regs on my first table, but as people got knocked out or moved around there were still some 40% vpip passive fish. I think typically below the $109 level you'll have at least 1-2 fish on a 9-handed table, plus some regs are bad/spewy and others prone to tilt. So I think with a small field, the edge is there for sure - it's just so, so much easier to reach a FT when you have only 100-something players to get through not 1k+.