While the advice is as good as it gets, I've often wondered about the last statement. So I pointed the trusty SD9 into a dark corner and fired off two shots. The first at ISO400 with zero exposure comp and metered center-weighted to zero on a knot in the wall boarding. The second at ISO100, the only change made to the camera settings - hence the metering said -2EV. Both raw, FWIW.
Then converted to TIFF, using dcraw which automatically brightens by "stretching" the tones - effectively making darker images brighter. Thus no post-processing actually needed in the test, dcraw does it for you

.
The main point being made here is that, if only the ISO is changed, the sensor exposure does not change and one would expect identical noise results from images made equally bright. And so it turns out (ISO400 at left):
cheers,