I guess Emad's book contains something to reply your question.
1. larger W/L increase harmonics, the more harmonics, the more phase noise we have.
2. larger W/L increase Ctail, thus more upconversion of low frequency tail noise.
MIT published a new paper on it 2005 JSSCC Feb. You can go and have a look.
rfmems wrote on Dec 17th, 2006, 7:20am:Sheldon,
You can find this statement in "The designer's guide to high-purity oscillators" by Hegazi on page 57.
Although the deduction is somehow wrong and unclear, but this statement however fits my calculation too.
Best regards
Chen