George provided several handwritten comments that were addressed.
Below are additional comments on Version 3.0.
================================================================

First two points are mostly curiosity, so you may or may not
want to modify the paper for them.  Next two points are worth
thinking about.  If you want to discuss (esp last point) I'm
around and relatively available this week.

I'm curious about the types of IR spectra you get when you sort
the sample by IR/UV.  I think it would be fun to see what alpha
is for the various spectra in fig 9 (sorted by IR/UV).  Does it
support the broad correlation between  IR/vis and color? Do the
spectra stray away from the single-galaxy behavior?  Part of the
answer to this question is in fig 13, showing the 70/160 wedge,
and that ratio being almost constant on average as IR/UV varies,
but i'm curious about the other wavelengths and what they do,
since the plots in fig 9 seems to show little change beyond 24um
but lots of change shortwards of that.

Similarly, I'm curious to find out whether the two eigenvector
spectra in fig 10 have any relevance to real IR spectra of
individual galaxies.  To my eye, they look vaguely like warm
and cool spectra, but that could be completely off.  There's
no plot that picks up that decomposition.

On the clumpiness/condensation business, this is still looking
very good.  My only concern is that these estimators will have
very different significance as you vary the distance over the
sample.  Let me suggest that fig 16 somehow add an indication
of distance, even if it's something like <10Mpc and >30Mpc.
I'm hoping it'll strengthen the case for resolution being a
meaningful discriminant.

--The figure now reflects distance information.  No trend with
  distance is observed.

I don'y buy your explanation of fig 17 spiral end.  I'm not
sure I have an alternative interpretation.  I'm worried the
increase in SSFR is driven mostly by an increase in UV, 
which would drive the apparent trend you see in the plot.

--After input from several others including David H., I have 
  changed the explanation.

I just remembered that Helene in her NGC300 paper combining
Spitzer and GALEX data found an interesting and intriguing
dependence of the UV extinction law on geometry.  Not entirely
sure if relevant to this paper, but take a look and see if
there is any potential connection.

--I have added a reference to Roussel et al. 2005, since their
  Argument #2 for increased FUV/TIR with galacto-centric 
  distance is similar to what I employ.
  
=============================================================
sec 5.2:  in middle of 1st para there is sentence "Put another way, IR/UV 
increases as redder. older stars increasingly dominate".
I found this sentence a non-sequitur, esp in light of the later discussion
in paper.  I think you can already here talk about this being driven by
the disk properties in terms of geometry in SFRegions, implying that
primary factor driving the change is porosity or covering factor in SFRegions.

--Good point.  I have modified the text.

sec 5.3, near end of 2nd para, you give a reason for why upper-left of
fig 13 is empty, which begs the question of why lower-left.
Also, in giving this explanation, i'd suggest you stick to direct physical
arguments, ie instead of "dusty regions are associated with vigorous SF,
which in turn implies warm dust", just say "high opacity implies higher
density dust, closer to heating sources, therefore warmer dust".
You comeback to this later on in the paper, but might as well do it here.

--OK, fixed your latter concern.  As for former, I don't have a decent
  explanation as for why we don't encounter cool, optically-thin galaxies.

sec 5.3 at end, you say distance is not a driver for fig 16.
If i refereed this paper i'd object: there is clearly a seondary
correlation with distance, so i'd suggest you amend with "distance
contributes but does not dominate as a driver for the effect"

--Thanks.  Wording altered.