It's much easier to hear a minor boost or cut using a low Q filter since it affects a broad range of frequencies. A high Q filter is less overtly audible because it affects a more narrow frequency range.
For general tonal shaping (making tonal changes intended to be audible) keep the Q rather low, and refer back and forth between whatever you are working on and an especially good sounding reference recording of the same genera. That will keep your ears honest, otherwise its easy to get led astray, or settle for something which is better than the original, but could be made significantly better still. A commercial recording or live release often makes for a superior reference than a recording of your own which you like the sound of. You may not get there entirely, but having that sound as reference makes a big difference. It may seem like one's memory of what it sounded like or some ephemeral idea of what it should sound like is enough to judge what is right, but its really easy to deceive one's self that way. Just use the reference recording as a tonal reference and ignore differences in level, loudness, dynamics and reverberant balance.
If targeting a specific problematic resonance, first find the center frequency of the sound in question with a high-Q filter set to an rather extreme boost, sweeping up and down until you home in on the specific frequency. This will sound terrible and is intended to. You're intentionally emphasizing the terribleness to precisely tune the filter to the middle of the effected frequency range. Once found, change the boost to a cut, attenuating instead of amplifying the offending sound, then play with the filter Q to find the most optimal width of the filter. By carefully adjusting the Q of the filter you can make the filter just wide enough to target most of the offending resonance without affecting too much of the frequency range to either side. You'll want to go back and forth between adjusting the Q setting and attenuation depth or level of the filter to find the most optimal setting because Q and level are so closely tied together in a psychoacoustic sense. The narrower the filter, the more you can cut without audibility and vice-versa.
Finding the best setting for that kind of problem reduction where the intention is not to hear what it is doing in the end result requires as much listening art as making good sounding broad tonal changes which are very much intended to be heard.
And doing that comparison check thing between your great sounding reference and what you are working on through more than one listening system (main stereo, car, headphones, etc) is also extremely useful for making sure you are doing more good than harm.