There are so many articles about this. I read one that had to do with writing down half and double of used sensitiviites, using them and then deducing by trial and error. Sounds good but I think with a little insight I have a creative way of getting the job done also. Muscle Memory is very important in getting consistant performance out of your mouse. I'm not certain if either low or high sens players have an advantage in this area. It would seem that more exaggerated movements would be easier for the brain to decifer and remember but I think our brains are pretty good to get the job done either way so maybe thats just another moot point.
It is my argument that the lowest speed you should choose is one that lets you track an enemy player that is circling you at the closest range possible and the highest sensitivity you should choose is one that still allows you to aim fast & concretely at stationary & moving enemy that are the farthest distance that you will encounter in your game. Between those points anything should be possible not just one "special number". Sensitivity isn't just about the speed of your mouse, it is affected by how fast you and other players move also. Thats how I came up with those to parameters.
Once you find a sensitivity setting you like based on the above and even if you choose another method you will find that somedays your aim is dead on, while other days you can't hit the broad side of a barn. I think this has to do with calibrating your hand to screen. Recovering the muscle memory and improving muscle memory is my goal. So these two things I'm sharing.
(these work best in spectator mode as long as you still have crosshair)
The first method is the most effective one. IF you've heard it before more power to you. 
Pick a point as close as possible and circle left and right while tracking the point with your crosshair.
Pick a point as far as possible and do the same. People with really high sensitivities (2.5inches =360) should appreciate this one the most.
And of course practice everything in between.

This next method seems to be more discipline than it is a calibration.
Pick a point, point at it with crosshair, move crosshair moderately to the right/left then try to bring it back to the choosen point without under or over shooting target. The faster and larger distance you can accurately do this the better.

This is a formula that I dug out of the boomslang forums some time ago.

Sensitivity-(arctan is the inverse of tan) arctan(2/resolution) = how much 1 equals of 360. so if 1600x1200, arctan(2/1600)=0.071619687089448084462549233000973 so set sensitivity to 1 and m_yaw to 0.07161 or you could take the (arctan(2/resolution)/(M_yaw)= sensitivity; (arctan(2/1600)/(.022)=3.2554403222476402028431469545897 (arctan(2/1280)/(.022)=4.0692992106339409584998445186424 all this bs would = the maximum sensitivity allowable for pixel perfect aiming independent from the mouse. arctan(2/resolution)*tan(FOV/2) complete formula

Another problem that I have is when I'm nervous. The adrenalin really can mess up your aim sometimes and sometimes it seems to make you super focused. So a lower sensitivity makes jitters not so much a prob. But you guys may not be using as high of a sensitivity as I have. Right now I have lowered my sens but I am still in the high sens range of things.
It is my assumption that the best sensitivity is the slowest one you can find that allows you to still track a target circling you at the closest range possible. For me this number is way lower than settings that Ive been normally using so to me it feels slow but is still what people call high sensitivity.
Two things that are true is that to a certain point reactivity can be made up for in lower sensitivities by moving your hand physically faster. On the flip side, to a certain point, control and precision can be made for in higher sensitivities by increasing dpi.

This article is not law, its just my way at the moment until I come up with something better.
Please leave your thoughts and better yet add to. I am always trying to find new ways of doing things.

Don't buy too much into the whole formula BS it just looks cool I guess. Maybe it gives insight on how technical you could make things but I don't think it needs to get so deep. The best way to judge whats going on is inches per 360.
Personally i've stopped adjusting the m_yaw because I can't figure out what exactly I'm doing/adjusting I'm essence but if m_yaw affects potential resolution a lower m_yaw would always better...I think...especially If you can't adjust the sensitivity setting lower than the m_yaw. Pretty sure thats right.
In CS 1.6 I cant adjsut sensitivity lower than 1 so I had to lower m_yaw to get a lower sensitivity. Just so happens that I cant go below .022 with m_yaw so sens 1 m_yaw .022 is the lowest possible sensitivity. Right now Im using a sensitivity of 1.255.