Dear Judith, 
Please find below the response to the referee report for paper 
MS# 70721 by Amaya Moro-Martin et al. entitled: "Are Debris 
Disks and Massive Planets Correlated?" 

Thanks for your consideration and thanks to the referee 
for very useful comments and his/her careful reading of 
the manuscript.

Cheers,
Amaya
______________________________

REFEREE:

  The main result of this paper - that there is no significant
correlation between planet and debris disk detections - is presented
concisely and clearly.  The implications are important for those
interested in extrasolar planets, debris disks, and the formation of
planetary systems.  With some clarification of the survival
analysis, I recommend this paper for publication.

  While the disk-planet correlation is the main subject of the paper,
the detection of a new debris disk + giant planet system (HD 38529) is
interesting in its own right, and has received a prominent place in
the abstract and summary.  I would suggest, however, that further
attention be given to this system, in light of the IRS spectrum.
There is currently very little discussion of this spectrum, mainly
just a passing mention of 0.7-sigma excess at 33um (in section 3).
I have not seen the spectrum in detail (and I would recommend adding
it as a new figure) but from the derived photometry in Table 3
it appears that the excess is much more significant.  It is important
to note that the claimed 6% calibration uncertainty for the IRS
photometry is for the overall spectrum, not for individual wavelengths
relative to each other.  As such, smaller deviations from the star's
wavelength^-2 flux can be detected.  Looking at Table 3, the slope
from 13um to 24um appears photospheric.  The 33um point, meanwhile,
is at least 10% higher than a fit to these points, a very significant
increase that I would think would stand out very prominently in a plot
of observed/model flux (again, a figure would help here).
  The significance of IRS excess depends on the data analysis.
Please indicate that the spectra have not been pinned to the stellar
photosphere and discuss how the various modules are spliced together.
  If the IRS excess is truly significant, additional analysis of the
orbiting dust is warranted, such as an improved fit to the dust
temperature and better estimates of the dust location, mass, etc.

RESPONSE:

The paper under consideration was originally part of a larger paper
that included the debris disk planet correlation plus a detailed 
analysis of the HD38529 system. For clarity, and because we considered 
that both topics were interesting in their own right, we decided to 
split the paper in two. This is the reason why the results for HD38529
are not discussed any further. To inform the reader that these results
are presented elsewhere, we have added a reference to the HD38529 paper
at the end of the introduction.  In this paper, expected to be submitted 
within the next 3 weeks, we model the disk SED, estimating the possible 
location of the emitting dust, and we model (analytically and 
numerically) the dynamical evolution of potential dust-producing 
planetesimals in the presence of the known two planets, trying 
to give a self consistent picture of this interesting system. 
I appreciate very much the above comments by the referee and I will 
introduce them in the HD38529 paper before submission.

_____________________

REFEREE:

  The most significant comments I have are for the survival analysis.
First, please describe the method a bit in section 4.2.
At a minimum, indicate that 1) this analysis is ideal for data sets
containing upper limits, but 2) it requires some assumptions for the
underlying distribution.  Explain that the three tests considered
(Gehan, logrank, Peto-Prentice) are based on different assumptions for
the flux distribution.  Is there any reason to think that one of these
tests would be most reliable for this particular dataset?

RESPONSE:

To address this point We have added the following to the survival analysis
section (now section 4.3):

"The planet sample and the control sample are dominated by upper limits, therefore, 
the K-S test is not sufficient to assess the probability that they could
have been drawn from the same parent distribution.
To extract the maximum amount of information from the non-detections
it is necessary to use survival analysis methods, which make certain
assumptions about the underlying distributions.  

(...)

As discussed in  Feigelson \& Nelson (\citeyear{feig85}), the logrank test is 
more sensitive to differences at low values of the variable under consideration
(i.e. near the upper limits), while the Gehan test is more sensitive to 
differences at the high end (i.e. in the detections). The Peto-Prentice test is 
preferred when the upper limits dominate and the sizes of the samples to be compared
differ (as it is our case)." 


____________________

REFEREE:

  I strongly question the use of 70um excess flux (in units of mJy) as
the variable in consideration.  Wouldn't it be more useful to consider
the excess luminosity (i.e. L_dust/L_*) rather than flux?  Flux will
presumably be most highly correlated with distance, rather than other
variables such as planets.  In several places (4.2 and the conclusion)
it is claimed that there is no correlation with excess luminosity,
when in fact the test uses flux.

RESPONSE:

Thanks for pointing that out. We have changed the survival analysis 
so that it now uses Ldust/L* as a variable, instead of the excess flux
in mJy. The probabilities have changed, but the conclusion 
is still the same (no evidence of correlation). We have modified the
the survival analysis section to reflect this change. 

________________

REFEREE:

  Given that the strength of the survival analysis is in dealing with
upper limits, please indicate what upper limits are used - 1-sigma or
3-sigma; observed plus noise or simply noise.

RESPONSE:

It's 3-sigma, as it is now indicated in section 4. 

_________________

REFEREE:

  Again, given the method's ability to handle different detection
limits, why were the two samples, FEPS & Bryden, only considered
separately?   You're explicitly taking into account any selection
effects based on distance and background noise, so can't the two
samples be combined into one?

RESPONSE:

The stars in Bryden's sample are sistematically closer than those
in the FEPS sample, that's why we originally kept both samples 
separate. In any case, to check if it would make any difference, we've 
repeated the survival analysis (with the excess luminosity) with both 
samples together and the results are now mentioned in the survival
analysis section. The conclusions don't change.

_________________

REFEREE:
Introduction:
  Is there a reference for the mass of the Kuiper Belt?

RESPONSE:
The mass estimate has been updated in the intro. Ref. are
Bernstein et al. 2004 and Luu and Jewitt 2002. 
_________________

REFEREE:
Section 2:
  Uncertainties for the three instruments are quoted from various
sources.  It would be informative to compare these uncertainties
with the dispersion within your own data relative to model values
(for IRS-13,24,33 and MIPS-24 in particular).
  How are the MIPS-70 uncertainties in Table 3 calculated?

RESPONSE:
We have added to section 2 a description of how the uncertainties
for the 70 micron fluxes were calculated. Thanks for pointing out
that this piece of information was missing. This is what we have
added: 

"The photometry uncertainty is given by $\sigma$ = $\Omega$$\sigma$$_{sky}$
($\it{N_{corr}}$)$^{1/2}$$\eta$$_{sky}$$\eta$$_{corr}$(1.0+$\it{N_{ap}/N_{sky}}$)$^{1/2}$, 
where $\sigma$$_{sky}$ is the standard deviation in the sky annulus surface 
brightness,  $\Omega$ is the pixel solid angle, $\it{N_{sky}}$ and $\it{N_{ap}}$
are the number of pixels in the sky annulus and in the aperture, and 
$\eta$$_{sky}$ and $\eta$$_{corr}$ are correction factors that account for 
the presence on the mosaic of non-uniform noise and of correlated noise, 
respectively. We used $\eta$$_{sky}$ = 2.5 and  $\eta$$_{corr}$ = 1.40
(see full description in Carperter et al. in preparation)."

A very detailed description of the FEPS data (including what
the referee suggests) will soon appear in Carpenter et al.  
This is why we chose not to present this same analysis here. 
We refer to Carpenter's paper through out the section. 
__________________

REFEREE:
Section 4:
  The second paragraph (on the color-color diagram) seems misplaced.
Should it go in section 3?

RESPONSE:
Agreed. It's now in the first paragraph of Section 3.
__________________

REFEREE:
Section 4.1:
  The K-S test shows that the control and planet samples are
consistent with each other in terms of distance and IR background.
Why was age, which was discussed at length in the first half of the
paragraph, left out in the second half?  Age is certainly the most
important variable to consider - survival analysis naturally takes
into account different distances and backgrounds, but not ages.
  A K-S test for age must also be included.  Also, please list the
number of stars in the two rough age bins, <1 Gyr and >1 Gyr.

RESPONSE:
The K-S test for age yield a very low probability Prob(D>obs)~1e-5, 
i.e. in terms of stellar age the planet and control sample have NOT 
been drawn from the same distribution (planet stars tend to be older). 
This would be a problem if there were a correlation between the 70 
micron excess and the age of the system for the ages under consideration. 
There is indeed a correlation for stars <100 Myr, but our control sample
is not including those younger bins, only stars older than 300 Myr. 
Observations indicate that for these older stars there is no strong 
correlation between 70 micron excess and age (Bryden et al. 2006 and 
also Hillenbrand et al. in preparation - the later are results from FEPS), 
and based on these results, we decided that it was reasonable to compare 
both samples, even though they differ in age. To clarify this point we have 
added this to section 4.1:

"Note that the K-S test for age yield a much lower probability 
($\it{Probability(D>observed)}$ $\sim$10$^{-5}$), i.e. both samples are likely not drawn
from the same distribution in terms of age. However, given that the observations indicate
that for the ages under consideration ($>$ 300 Myr, with approximately half
of the stars having ages $>$ 1 Gyr) there is no correlation between the 
70 $\mu$~excess and the stellar age, we do not expect to introduce any significant bias 
by comparing both samples (but keep in mind that the validity 
of the comparison relies on the observed lack of correlation with age). "

_________________

REFEREE:
Section 4.2 (Survival Analysis)
  The second paragraph has nothing to do with survival analysis.
Maybe it should go in 4.1?

RESPONSE:
Yes, it was out of place. Section 4.1 describes the sample selection 
so it probably doesn't belong there either. I've added a new small 
subsection "4.2 Frequency of debris disks". And for consistency, I've
changed the title of old subsection 4.2 (now 4.3) to
"4.3 Fractional excess luminosity: survival analysis". This way, 
the sample selection criteria and the results are in different
subsections. 
_________________

REFEREE:
Section 5.2, 3rd paragraph
  The caption to figure 3 states that the model might be consistent
with the data if the model is allowed to float by a factor of 10.
  A stronger and perhaps more interesting statement would be to rule
out the model based on its predicted time dependence, which is not
observed (also see Najita & Williams 2005 who show that the model's
predicted evolution of disk radii is not seen in observations).

RESPONSE:
Agreed. We have added the following sentences to section 5.2. 

"However, this cannot be the only mechanism because if it were to dominate debris production
one would expect to see the dust temperature to be correlated with age and this
trend has not been observed (Najita \& Williams~\citeyear{naji05}). Similarly, the 
observations in Fig. 3 could not confirm the time dependence of the fractional 
70 $\mu$m excess emission predicted by the models."
 
_________________

REFEREE:
Section 5.2, 4th paragraph
  Exactly what type of planets does this 12% frequency consider?
The text refers to gas giants with mass < 13MJup, but certainly this
does not refer to Neptune-mass planets, for which little is known.

RESPONSE:

Yes, it cannot refer to Neptune-like planets. We've added a few 
clarifications to that paragraph about those estimates:

"Firstly, debris disks are more common than massive planets: 
it is found that $>$7\% of stars have giant planets with M$<$13 M$_{Jup}$ 
and semimajor axis within 5 AU, but this is a lower limit because the duration 
of the surveys (6--8 years) limits the ability to detect planets between 3 AU 
and 5 AU. The expected frequency of gas giant planets increases 
to $\sim$12\% when RV surveys are extrapolated to 20 AU (Marcy et al.~\citeyear{marc05}), 
with the distribution of planets following d$\it{N}$/d$\it{M}$ 
$\propto$$\it{M}$$^{-1.05}$ from M$_{Saturn}$ to 
10 M$_{Jup}$ (the surveys are incomplete at smaller masses)."
 
_________________

REFEREE:
Section 5.2, 5th paragraph
  Besides the Greaves 2006 reference, there are several papers looking
at the correlation with metallicity based on more up-to-date (Spitzer)
data (e.g. Beichman 2005, Bryden 2006).

RESPONSE:
Of course. Those two references have been added. 

_________________

REFEREE:
Table 1:
  Although it is hard to determine statistical trends with just a
single detection, it may be worth noting that the one star with
excess, HD 38529, is the most evolved, most luminous, and most massive
star in the sample.  Particular so, given theories that predict
melting of Kuiper Belt material as a star evolves (Jura 2004).
At the very least, its unusual placement in the H-R diagram
(well above the main sequence) confirms its old age (>Gyr).

RESPONSE:
Thanks for pointing that out. We have added the referee's comment to
to end of the second paragraph of section 3 (where the HD 38529 
detection is discussed). 

"Even though it is difficult to
identify statistical trends  from one detection, it is interesting to note 
that HD 38529 is the most luminous, most massive, and most evolved of the planet
bearing stars in Table 1. Assuming V = 5.95 (Johnson 1966), a Hipparcos distance
of 42 pc and no reddening, the object has an absolute visual magnitude of M$_{v}$ = 
2.81 and Log(L/L$_{Sun}$) = 0.82, putting the star on the Hertzsprung gap, so 
it is clearly post-main sequence."
 
________________

REFEREE:
Table 3:
  Perhaps note that calibration uncertainties are not included in the
error bars.

RESPONSE:
Done. This is clarified now in the table caption.

_______________

REFEREE:
Figure 1:
  Having excess shown as a subpanel below each SED is a nice way of
presenting this type of data that I haven't seen before.

RESPONSE:
Thanks.
_______________

REFEREE:
  Finally, as to the main conclusion of the paper (no correlation
between planets and IR excess), is it possible to place a quantitative
limit on this lack of correlation?  For example, can you state that
the frequency of debris around planet-bearing stars must be within a
factor of two of control sample's?

RESPONSE:
We have the following detection frequencies:
1/9 for planet stars vs. 9/99 for control sample (FEPS)
1/11 for planet stars vs. 7/69 for control sample (Bryden). 
If we assume that the error in the number of stars with excesses
is ~ sqrt(N), then we could state that the frequency of debris around
a planet-baring star is within a factor of 3 of the control sample. 
It's a crude estimate because we are dealing with very small number 
statistics. We've added this to section 4.2. 

_______________


Typos:

>>"`scale" --> "scale" in section 2, paragraph 2
I didn't find this one...

>>"asses" --> "assess" in section 4.1
Done.

>>"cannot not" --> "cannot" in the appendix
Done. 

>>The lower left panel of Figure 2 appears to have an error.
>>Where does the point with F24/F8 = 0.12 come from?

Thanks for pointing that out. That point was HD 80606. 
The reason why it did  not appear in the panel above 
(F33/F24 vs F24/F8) was because its IRS spectrum
is very noisy beyond 30 microns, making the 33 micron 
IRS synthetic photometric point unreliable (this source was 
flagged by the SSC as non-nominal possibly because of a peak-up 
failure). However, in checking this out I've realized that there 
was a small inconsistency between the 24 micron values that I was
using to create the color-color plot  and the 24 micron values 
listed in table 3, the latter coming from our latest data 
reductions with improved 24 micron photometry. The issue was 
that I had forgotten to update the color color plot with these 
new values. The new manuscript now includes the updated 
color-color plot, including HD 80606 in both panels. I have 
added a note about this star in the table caption explaining
why it's an outlier in the upper left panel. Thanks 
again for pointing out this inconsistency. 

>>"could consistent" --> "could be consistent" in Fig. 3 caption
Done.

*************************************************