bitflipper
Plenty to chew on so far, especially tlw's excellent comments. As you can see, it's all a bit more complicated than it seems it should be - a common theme in this arena, in which so many aspects are unexpectedly counter- or non-intuitive. 
 
Given the good advice already offered above, I'll add just one practical observation: your car is a terrible place to judge mixes.
 
I don't care whether you have a high-end aftermarket stereo system or a stock factory installation in your car. It's too small a space, too many odd reflections, and too many resonant cavities to allow an honest representation of a recording. Even an expensive system will play tricks on you, usually adding an unadvertised low-frequency boost in an attempt to counter the acoustical limitations of a car's interior.
 
Add to that the fact that you are not monitoring bass accurately in your studio and you have a deadly combination that's guaranteed to bring endless frustration. The most common symptom will be excessive bass when you listen in the car. You'll go back to the studio and try to fix it, but the only way it sounds OK in the car is to make it sound thin and nasty in the studio.
 
I burned through a LOT of CDs chasing my tail that way, until I took my mixes to a professional studio and judged them there. That's where I was struck by an epiphany: my studio monitors were lying, my living room stereo was lying, and my car's stereo was lying. They were all lying to me, but each in a different way, making it impossible to create a mix that sounded good on all three.
 
The ultimate solution was to create a neutral monitoring environment in the studio. That meant a serious investment in quality speakers, acoustic treatments, and objective measurements to identify where my monitoring weaknesses lay. Only then could I rely on what I was hearing.
 
Mixes still sound a bit boomy in the car, but I now know that's the fault of the car, not my mix. With my ears better attuned, I now realize that even well-made commercial CDs are a bit boomy in the car. But they fall within an acceptable range of boominess because the mastering engineer had tuned the mix to a neutral reference.
 
You certainly don't want to hear that the solution to your dilemma is spending thousands of dollars and weeks studying up on acoustics. Unfortunately, that's where you're ultimately headed if you're serious about this stuff. The good news is there are things you can do now that cost nothing other than your time.
 
Your friend in this endeavor is the spectrum analyzer. There is an excellent one called Voxengo SPAN that may be downloaded for free. Start there.
 
With SPAN on your master bus, rip some tracks from your favorite commercial CDs, import them into SONAR and examine their spectral breakdowns. You will quickly see commonalities between well-made records in the way frequencies are distributed. This doesn't mean you should necessarily force your mixes to mimic them exactly, but they will give you a guide so you'll at least know when you're inside or outside the ballpark. 
 
Now, old hands will always say "trust your ears" and "don't rely on visual aids". Ignore them. At this point you cannot trust your ears and you cannot trust your monitoring environment. Visual aids are your only option. At least, until you've got a couple extra grand in your pocket.

Good stuff, Bit. I was actually hoping you'd chime in on this. Can you tell me what the difference is between the Voxengo and the Spectrum Analyzer in the Pro Channel EQ?

Tosmurf
I have learned i really nice "trick" from Grammy Winner Jacquire King.
He wrote this in an article years ago. Its about getting your bass allways to the right volume. Dosent matter where you mix (other room other sound...).
 
I personally use this in every mix i do. 
 
In short its seting up your mix bus like this.
 
MY mixBass for example is  
EQ plugin
Mix Bus Compressor
EQ plugin
Tape emulation (vu meter) 
 
The trick now is that a VU meter is really sensitiv to low frequencies not to high. So if i start a mix i start with my kick drum volumen- It should peaki at -3dbvu (tape emulation VU Meter).
Things to know if you double an instrument the dbvu goes up by 3dbvu. So bass and kick have nearly the same frequencies. They are in the same ballpark. 
So my kick peaks around -3dbvu. If i bring the bass up -  bass+kick should peak at 0 dbvu.
Now build your mix around this gainstaging. And your bass should be allways in the ball park.
 
There was a youtube video horrible long but he explains it. I ll search it and post the link here.
 
EDIT
www.youtube.com/watch?v=lF2oNsiZWjc
There you go horrible long but worth looking expecially the bass volume trick ist woth this 50 minutes ;)

Thanks Smurf. I'll see what I can do with that formula. I've actually been thinking of VU meters due to some previous posts from Bitflipper. Didn't even think of the tape emulator having one. I was gonna go buy a VU plug-in. Great information! Thanks again.
 
~Jeff

On the subject of room treatments, commercial products can be very expensive, and some aren't as effective as you might hope. Acoustic tiles a couple of inches thick, for example, will reduce high frequency echoes up around 10KHz+ a bit, but will do nothing for the mids or bass.

It's quite possible to do a lot for a room relatively cheaply. Thick quilts are useful, as is padded furniture. A simple and pretty effective all frequency trap can be made out of a few square feet of heavy rockwool insulation suspended a fre inches below the ceiling or off a wall, or placed across the room corners, wall/ceiling joint etc.

Sound diffusers can be as useful as sound absorbers, and can be made easily, or even bookshelves containing books of varying depths can work quite well. There are loads of examples and DIY projects on-line, try searching for "studio sound diffuser".

Can you tell me what the difference is between the Voxengo and the Spectrum Analyzer in the Pro Channel EQ?

SPAN has far more options, such as slower averaging modes that are more helpful for mastering, the ability to freeze the display for taking screenshots and making comparisons, a correlation meter, clipping sensor, Mid/Side mode and a resizeable display. There are lots of spectrum analyzers out there, including some much more sophisticated than SPAN, but they're not free.

A lot of effort and good information given by some great people here. It really is a matter of wading through the various methods and finding which ones help solve your problem. I am in complete agreement with those that say your car is not the best place to judge any mix. Although, if you get a mix that works well in your car you will no doubt be on the right track.
I've always had trouble making mixes that translate well to my motorbike.

There has been a great deal of good advice, but there have also been a couple of "old wive's tales" that really need to be put to rest...
 
the decibBel is, as stated, 1/10 of a Bel, which was a unit-less measurement introduced by the Bell System to make transmission loss easier to work with. The original unit of measure (really testing my memory here) was Mile of Acceptable Cable. No, I'm not kidding!

None of which matters... what matters is that the dB is a ratio, and specifically, a power ratio. By definition it is an rms measurement - it can not be properly used to represent peak levels in analog measurements (more on this in a moment.)
 
rms means root-mean-squared, another handy little mathematical too that lets us calculate the effective power in a periodic electrical (or magnetic) wave. It is equal to the DC voltage that would produce the heat (dissipated power) in a pure resistance. All if which is important if you are a measurement geek, for us the things that matter are that it is used to measure the effective power in a periodic wave, and there is very little music that can be represented by a periodic wave, but all music is made up of periodic waves.
 
The dB ratio is equal to 10 times the log of the ratio of two power levels. If you know Ohm's Law you can manipulate this a little bit and come up with the forumla of dB = 20 times the log of the ratio of two amplitudes or levels (or in our case, voltages.) And for the curious, the forumula for the Bel is the log of the ratio of two power levels. But if you were curious you probably already guessed as much.

dB becomes much more useful when you replace the bottom half of the ratio with a known value.
 
In the beginning there was dBm0 which meant the ratio of the power in a circuit with respect to 1 mW across 600 ohms. This is still quite useful in telephony and RF, but it doesn't help us much, matched power transmission went out of style in the  early 1970s.

We are much more interested in dB scales that reference voltages. The first such scale used the equivalent voltage one would measure in that 1 mW signal across 600 Ohms, or approximately 0.7746Vrms. We refer to that today as 0 dBu, and there is no requirement for impedance, it is simply a voltage ratio.
 
It remains the basis of the "professional" audio interface. +4 dBu has become the reference level for a great deal of equipment, most often balanced inputs and outputs. That is not a requirement.

As consumer "Hi-Fi" gear grew in popularity the balanced interface was replaced with a single-ended interface, and the nominal operating level was changed from +4 dBu to -10 dBV. The dBV scale uses as its reference 1 volt. A lot easier to do the math, but it never really took off in professional circles. (Even today many broadcasters think +4 dBu is a silly standard, they prefer +8 dBu.)

Wandering off just a little bit - but there is a good reason - lets convert our nominal levels to their voltages:
 
+4 dBu = 1.227Vrms = 3.472Vp-p (and approximately 1.78 dBV)
-10 dBV = 0.316Vrms = 0.894Vp-p (and approximately -7.78dBu)
The two nominal levels are approximately 11 dB apart - just for grins!
 
OK, so what about this pesky digital audio?

All of the sudden we need to be able to think in terms of peak voltage if for no other reason that we do not wish to clip our converters. From this was born the dBFS scale, and it is slightly different because the reference is the maximum, there is no way - by definition - to go over 0 dBFS.

Which, as it turns out, really doesn't matter a whole lot either!
 
What matters, and what many folks have been trying to say, is that the relationship between 0 dBFS and the maximum analog voltage you expect to need is the important thing.

For example, lets say I have a microphone preamplifier that is capable of putting out +24 dBu (about 12.7Vrms or about 34 Vp-p). If we set 0 dBFS to equal +24 dBu then our nominal operating level will be about -20 dBFS if we are thinking in terms of +4 dBu analog audio.

That paragraph may trip up a lot of folks so I'm going to leave the math portion of this post for now - please feel free to offer questions, corrections, etc...

So when audio became a thing some bright folks noticed that neither a peak reading nor an RMS reading meter accurately predicted squat - you couldn't use it to manage modulation in your transmitter, and you couldn't use it as a benchmark about loudness. And that's what we care about!

So a new measurement was developed - the Volume Unit. You've seen (I hope) VU meters on audio gear. What makes a VU meter special is the ballistics, or how quickly it reacts to positive and negative going changes. It is WAY to slow to show a typical musical peak. But once you get accustomed to it you can get a pretty good sense of relative loudness.
In pro audio 0 VU = +4 dBu, in consumer gear 0 VU usually = -10 dBV. AHA, that dB stuff does matter eh?

No one that I am aware of has come up with a digital meter that mimics the ballistics of an analog meter. The Durrough meter plugin from Waves comes REALLY close. There are others too, I'm sure, but  I stopped looking a long time ago and just use analog meters<G>!
 
The meters in most software platforms do a really good job of reporting peak levels, and for the most part they do a decent job of reporting RMS levels. They just don't rise and fall with the same timing as the old mechanical meters.

Which may well be a good thing! In the  old days we didn't care much about peaks because there were about a dozen things in the audio path that would squish them anyway - especially as we got closer to the power supply voltage.
 
Not so in the digital realm - if  the analog circuit doesn't do too much damage the digital part will faithfully reproduce what ever we feed it - or create. So we really need peak responding meters in our software, we just  need to remember that they do not give us much of an idea of loudness. (A really short peak can reach the power supply rail and be effectively inaudible. How's that for trickery?)
 
Last thing for tonight (my finger are getting tired) - channels do not add algebraically unless the signals are identical land in phase. Say what?

Well if you've followed along this far you'd know that if I add two identical signals together I'll end up  with a signal that is 6 dB louder. But individual tracks are not identical (who really wants to listen to 16 tracks of 1 kHz Sine wave anyway?)

So they don't add up, and in fact can be more - or less - than the 6 dB difference one might expect.

By the way, using sine waves is an old trick and very handy - if I apply a -6 dBu sine wave to two tracks and then send them to a buss the buss will measure 0 dBu. It's a terrific trick for balancing a stereo circuit, or two monitors. And we used to use tapes with a single track across two head gaps to align tape machines using the same principle.

I know that's a lot, and I know there is a lot more, but I thought  this might be a good starting point.

T-RackS CS is a FREEbie and includes a really good meter with perceived loudness
 
I really this meter and the Waves Dorrough meter, but that one isn't a freebie.

Great advice from many here. I would only add that anyone mixing in less than ideal spaces, or who would like to test their mixes in various virtual environments should check out IK Multimedia ARC2. I have found it far more useful than I ever imagined it would be.
 
http://www.ikmultimedia.com/products/arc/