I'd been faced with the same problem, lowering my paperdoll crit when attacking high lvl target.
I'm 75lvl, testing on 80lvl target dummy. So it's 5lvl higher me (Defense Skill - Weapon Skill = 25)

I did several tests varying paperdoll crit (fixed paperdoll hit, fixed talents)
I attacked target dummy from the back (for excluding blocks from measurement)

Screen of recount windows here ( be careful, it's ruRU locale, lots of cyrillic symbols):

My results:
ppd - paperdoll
nb - number

I apply linear regression to these data. Tried to find Crit% = A * Crit(ppd) + B;
I used regression tool embedded in the MS Excel.
Result:
So: Crit% = 1.002 * Crit(pps) - 10.8;
I dont think that real crit % could be multiplyed to the number greater than 1.000. It would be really strange.
As result i can only suppose that A=1 and Crit% = Crit(ppd) - B

So try to calculate B:
depr - depreciate

So crit reducing is 10,7503% when 75lvl attacking 80lvl.
If i suppose that this number grows linear with weapon skill decreasing we can get following formula:
CritReducing% = 0,4300% * (DefenseSkill - WeaponSkill);

P.S.
Note. Results for crit% are diferent for cases when we attack dummy from front ( it blocks) and from back (it doesn't block).
For examle, i'd done 2 tests with 20,78% paperdoll crit:
- [From front, blocked about 5%] My critchance was 9.16% with confidence interval [8,64%; 9,73%]
- [From back, blocked 0%] My critchance was 10.13%% with confidence interval [9,19%; 11,18%]
As you can see 9.16% doesn't match 2nd confidence interval, and 10.13% doesn't match 1st.
I calculated intervals for 95% confidence probability.

(I did this testing a few weeks ago, and finally have the time to write it up.)

Hypothesis
After looking over the tests posted in this thread (up to about 12/14/2008), and reading Vulajin's test results in the rogue thread, I began to wonder if ranged damage had a different rate of crit suppression vs bosses than melee. Vulajin found a crit suppression rate of 4.8% for melee, while the tests posted by hunters were showing an average of 5.3%. Could it be that ranged had an additional 0.5% of crit suppression to "balance" the fact that ranged gets no glancing blows?


Short answer
The short answer is no, ranged does not have a higher rate of crit suppression. Although negative results are not as exciting as positive results, they can give us confidence in our conclusions. Read on if you wish to see what I did to reach that conclusion...

Long answer
Since many of the hunter results I was looking at were done to test hit rating, and thus were done with varying levels of hit rating with and without Focused Aim (see my previous posts in this thread), I wanted to start with a clean hit-capped sample.

All my testing was done on the Darnassus test dummies. (I don't know about other servers, but on my server, if you want to go about your business undisturbed, Darnassus is the place to go. It is practically deserted except for deer, wisps, and the occasional low-level druid looking for handouts or advice. But I digress.)

Here is the result of a fairly long hit-capped test:
Heroic Test Dummy -- 5001 Autoshots
0/3 FA
5/5 Lethal Shots
266 (8.11%) hit rating from gear

22.12% Paper Doll crit
16.8% Actual crit
5.32% Crit depression
Screenshot

Once again the 5.32% figure showed up, so I decided to try a new test using Vulajin's methodology.

I juggled my gear to give myself a 5.32% crit rating, and started shooting to see if I would get any crits. I did the test both hit-capped and hit-naked, with no spec. As you can see, it took only 1000 shots in each case to generate 4 crits, which is in line with a 4.8% crit supression. 4.8 crits per 1000 would have been perfect results, but I did not continue the tests as my hypothesis of a 5.3% crit suppression for range had already been disproved.

Heroic Test Dummy -- 1003 Autoshots
0/0/0 spec
278 hit rating from gear

5.32% Paper Doll crit
0.4% Actual crit
4.92% Crit depression
Screenshot


Heroic Test Dummy -- 1010 Autoshots
0/0/0 spec
0 hit rating from gear

5.32% Paper Doll crit
0.4% Actual crit
4.92% Crit depression
Screenshot



Therefore I conclude that we can assume the 4.8% base crit suppression vs bosses is also valid for ranged, at least with no spec and when hit-capped.


Suggestions for further research:
I still wonder why the average of 5.3% crit suppression shows up in tests when the tester has talent points. My own previous test showed that getting part of one's crit from Lethal Shots did not substantially affect crit rate. Larger samples may show a slight affect. Thayer showed that getting some of one's hit rating from Focused Aim does seem to affect crit rate. However, I have done all my tests without Focused Aim, so there must be other factors as well. Or is it merely the size of our samples not being large enough to overcome the RNG?

I still find that even at 4000 - 5000 shots, RNG fluctuates as much as 0.8 percent in the last 1000 shots, so I think a larger test sample s needed.

That is why I haven't retested the effect of FA on crit to certainty, and I all I can say now is that there is evidence of it, since 2 percent after 5000 shots consistently can not be attributed to RNG. I have had a lot of stuff to do, so I will test FA soon, as this is a bug we need to get submitted to PTR if at all possible.

(I moved part of my data to this post because I have more test data to go with it, and I felt it would be clearer in its own post.)

In my results above, I was intrigued that I got the same number of crits in 1000 shots both hit-capped and hit-naked. Since these samples were so small, I then did some larger samples at 10.0% crit, hit-capped from gear vs hit-naked, no spec, to see if there was a difference. I then ran another test to try and confirm Thayer's data showing the possibility that using the talent Focused Aim was causing him to lose 2% crit above and beyond the documented 4.8% crit suppression on bosses.

As usual, all tests done on the Darnassus Heroic Test Dummies.


Heroic Test Dummy -- 5000 Autoshots
0/0/0 spec
272 hit rating from gear

10.00% Paper Doll crit
5.2% Actual crit
4.8% Crit depression
0.0% Misses
Screenshot


Heroic Test Dummy -- 7501 Autoshots (Round 1: 2500; Round 2: 5001)
0/0/0 spec
0 hit rating from gear

10.01% Paper Doll crit (10.00% in Round 1)
4.28% Actual crit
5.73% Crit depression
8.51% Misses
Screenshot Round 1
(Look at the chat log in this screenshot to see an example of the type of conversations one has with the Darnassus locals near the test dummies.)
Screenshot Round 2


On the hit-capped test, I got a perfect 4.8% crit suppression. The hit-naked test was cut short because I got called away to a guild heroic, so the apparent increased crit suppression could be due to randomness -- I'll go back and finish the test when the servers come up today and edit in the results.

EDIT: Since the numbers on the hit-naked test were looking a little weird, I ran an additional 5001 shots and combined the data from the two rounds to get a larger sample. In Round 1 I had 10.00% paper doll crit, but in Round 2 I have 10.01% -- it seems I got rid of a piece of gear since the first round and I couldn't make my gear come out to exactly 10.00% again.

Two things are worthy of note on the hit-naked test. One, my miss rate was higher than expected at 8.51% over the course of 7500 shots. Still within RNG from what Thayer has posted above, but a little surprising to me. Two, my crit rate was 4.28%, giving a crit suppression of 5.73%. This is nearly 1% (0.93%) worse than one would expect given the 4.8% crit suppression rate.

I ask any of you who are more adept with WoW statistics than I -- is this the rate that would be expected for the two-roll system with this level of crit?

I am now running a test with 5% hit from gear and 3/3 FA to see if I can replicate Thayer's results. If I can confirm the 2% crit loss using FA, that would be even worse for crit than having no hit rating at all!

Results of Focused Aim test:

Heroic Test Dummy -- 8000 Autoshots
0/3/0 spec 3/3 Focused Aim
165 (5.03%) hit rating from gear plus 3% from Focused Aim

10.01% Paper Doll crit
5.2% Actual crit
4.8% Crit depression
0.0% Misses
Screenshot

As you can see, I was not able to confirm a problem with Focused Aim; in fact, the results were exactly as expected with the 4.8% crit suppression against bosses. For the last 3000 shots of the test, the crit rate hovered between 5.3% and 5.1%, settling into 5.2% in the last 1000, so I do feel it is a reasonably accurate reflection. I have not run a test on a lvl 80 dummy; a longer test on one would probably be in order.

It should be noted that the ONLY talent I had when conducting this test was 3/3 Focused Aim; I couldn't see for sure, but in Thayer's screenshot, it appeared that he had 19 or fewer unspent talent points, indicating that he had many other talents in one of the trees besides the MM tree which is showing. Any of the tests results I have looked at in these forums which gave a crit suppression of more than about 4.8% (including my own tests) were conducted with partial or nearly complete specs. When I do a stripped down spec, the test results are as predicted. It may be Focused Aim interacting incorrectly with another talent or talents that is causing the problem, or as Thayer pointed out, we just need larger samples.

EDIT: Or I suppose it's possible that it was hot-fixed in the last week or two.

During today's tests on the ptr I have found that Wild quiver shots are not affected by lethal shot. At the moment I have less then 500 shots from Wild quiver, but it is a straight 5% crit reduction form my roughly 5k auto shot count.

Wild Quiver: (3/3)
You have a 10% chance to shoot an additional shot when doing damage with your auto shot, dealing 50% Nature damage. Wild Quiver consumes no ammo.

Technically this extra shot is from our ranged weapon so it should be affected by lethal shot, if I am wrong correct me before i do more tests.

I had enough points up in BM tree to get serpent's swiftness, and I avoided any abilities that effected my DPS directly, such as Imp AOTH, FI and such,putting all the points into my pet and defensive skills.

I wanted the ability to make the test go 20 percent faster.

you need to fall into the 6,000 - 10,000 shot range to generate consistent numbers.

it has already been proven that with 8 percent hit from gear, we always get a return 4.8 percent crit reduction. the main concern is of getting 5 percent from gear and 3 percent from Focused aim is generating 2 percent less crit, and this combined with the pet not getting the hit as well, resulting in a net DPS loss from going 3/3 FA.

I think it might be advisable to run a longer test before accepting that "getting 5 percent from gear and 3 percent from Focused aim is generating 2 percent less crit." Unless I missed something, the tests on which you based this finding consisted of 3000 shots on a heroic dummy and 2000 on a lvl 80 dummy.

I found no discrepancy when using 3/3 Focused Aim for 3% of my hit in an 8000 shot test. A longer test using the same spec you used for your shorter tests might turn up a faulty interaction between Focused Aim and some other talent, but I think we cannot blame Focused Aim alone for your result of 2% crit loss. If I had ended my test at 3000 shots it would have shown a 1% "increase" in crit from using Focused Aim; that is why I let it run so long.

The fact that the pet does not receive the 3% hit from Focused Aim, however, is a valid reason not to use this talent to reach hit cap if it can be helped.

I have started a new thread since the original title no longer applies to the discussion, and I do not want redundant post about things we have already proven.

The new thread is a discussion on the effects of Focused Aim on Crit Returns.

Focused Aim reducing Crit Return

i was also noticing earlier that my crit in recount was showing up about 5-6% less than it should.
though later on I seemed to have no problems. Id respec to 0/0/0 but I cant afford it at the moment.

Go ahead and look at the thread in my previous post. there has been a lot more information found in that one.