There are many different way you could set up a test depending how accurate you want your results to be. At the most basic level, yes, you just stick to your rotation and empty your mana pool. However, some people would prefer to remove some of the variables first so that as few as possible effect the results. For example, uneven LO distribution could sway the figures so you might want to take a spec without that to reduce the influence of RNG. The EO damage boost might also have a similar effect (imagine the difference between a test where every LvB had +10% damage vs one where none did). You might want to remove your other trinket so that it's procs do not influence the figures.

However, every change you make moves you further away from the 'reality' of your situation in a boss fight and you are already someway from it by not being fully raid buffed. Personally, I would keep your standard spec and gearing and just repeat the test many times to try and iron out RNG anomalies. Take a note each time of the crit %s on your spells and if any tests have abnormally high or low crit %s, throw them out. From each set of results, throw out the highest and lowest figures anyway so you are only looking at 'average' kind of figures - who knows what random occurence made them so high or so low? And of course make sure that nobody else is there debuffing the target dummy on some of the tests.

One other thing is that it might be worth stopping your rotation the first time you reach the end of it after getting below 2000 mana. There are 2 reasons for this. The first is that to be a really fair comparison you need to end all tests at the same point - a test that ends in a LvB will have slightly higher DPS than one that ends just before you are about to cast one. Also, by finishing at the end of your rotation, you ensure that the FS is not ticking as the ticks of the FS after you stop casting still get added to your numbers and lower the DPS.

As I suspect the results will come out fairly close, it will take quite a few tests before you can be sure you have found a reliable answer - be prepared to take some time over this!

Check the Elemental TTT, it's what I used to fix my shaman up: http://elitistjerks.com/f47/t20914-shaman_elemental/

The only difference I made was if I 'needed' a blue gem I would use the hit/mp5 one instead of the hit/stam one in the TTT.

In terms of the whole "waiting" discussion, things happened rather the other way around.

Back in the dim distant past (ie. WotLK beta) there was a lot of talk about ideal haste ratings, haste "dead zones" and waiting times. Things were obviously a lot different then (mainly due to no Shamanism etc.) but the concept of waiting to cast LvB in some situations seemed to be at least theoretically a good one.

More recently a lot of anecdotal evidence has built up that waiting is always bad. I want to be very careful here, as to be honest the theorycraft doesn't really back up the evidence. In theory waiting 0.2secs out of a rotation that takes 20secs+ is not a big deal at all, just a loss of <1% of DPS time. In reality it appears to knock a much, much bigger % off your DPS. The argument has been suggested that waiting is made worse by having nothing in the pre-cast system, and therefore you will be hit with the full effects of latency... but even that doesn't fully explain it to me.

So yes, I would always cast something - or even better check your rotation to try and minimize the awkward 0.2sec gaps. But I wouldn't say we fully understand the mechanics.

Pretty sure you want to keep Glyph of LB, but Glyph of ToW and Glyph of Lava are pretty close. I believe the conclusion was that Glyph of ToW was better until you get above a certain amount of spellpower (~2400 unbuffed I think?).