It's important to define the star ground point as a place that has no
connection to the chassis. That way you can check your work with
an ohmmeter to chassis, and when everything checks out, put that
final wire in. 

The last wire might be a diode bridge paralleled by a 10 ohm
resistor and a cap, too.
 
Obviously, the most critical stage is an input stage, and the closer
you go to the output stage, the bigger the signal level and the less
the milli- and micro-volt signals in the shared ground will be. 

So you can achieve some reduction by breaking either end out of
the shared-ground chassis. Break the small signal stages out of
chassis ground and they don't pick up the IR noise from the high
current stages that you've left in the chassis; Break the high current
stages out and the currents remaining in the chassis get a lot smaller.
The shared-resistance ground noise gets smaller either way. Is it
enough to leave the output stage currents in the chassis? For all but
the most critical situations, yeah, probably. 

Bias circuits are things that I consider sacred. Make a mistake in a
bias circuit and you get big problems. I'd recommend lavishing
great care on a bias circuit, and so while it may be perfectly find
grounded to the chassis, I'd star ground it on principle. It *is* an
alternate input to your output stage, so if there's hum or RF on your
chassis, it's conducted into the inputs of your output tubes. I'd star
the bias - unless actual listening says it works fine without it. 

 substantial ground buss... much easier to contend with than a true
star scheme. see it here, if you haven't already:

http://www.geocities.com/SunsetStrip/Garage/5701/preampmods.html

basically it's a tinned 12awg strip of bare wire running the length of
the chassis, supported by the center lugs of the 9-pin sockets and
the occasional terminal lug here and there.

works a charm!
  

The brass plate is actually a ground bus.....its the same thing as
using the chassis except that brass is a better conductor than steel.
The seemingly random grounds on Fenders are actually spots that
were chosen after much painstaking prototyping. If you look at the
way Fenders are grounded, the principle of keeping the low current
circuits at the far end of the amp away from the PT and the heavy
currents such as the first filter cap and the PTCT next to the PT are
used in these amps much the same way as in an amp that is star
grounded.

    I try to follow star grounding techniques, but sometimes the idea
of running a 16" length of wire around to the star ground point
seems a bit excessive. The post that points out that Fender's
grounding scheme was not as haphazard as it might appear is right
on the money... if you just ground everything to the chassis at
random you can get "eddy currents" (is that the right term?) flowing
all over the chassis, which could be picked up in the initial preamp
stages and amplified a few million times (well, a lot of that gain is
thrown out between the stages...