=========================================================================================
"Implementation of an F-statistic all-sky search for continuous
gravitational waves in Virgo VSR1 data" by Aasi, Junaid; Abbott,
Benjamin; Abbott, Richard; Abbott, Thomas; Abernathy, Matthew;
Accadia, Thimothee; Acernese, Fausto; Ackley, Kendall; Adams, Carl;
Adams, Thomas; Bejger, Micha³; Gonzalez, Gabriela; Krolak, Andrzej;
Riles, Keith; Vinet, Jean-Yves; Woan, Graham
Article reference: CQG-100417
Reply to referee reports:
Referee: 1
One minor comment is on the captions of figures: some figures appears colored
online, and do not fit with the explanations in captions.
R: We have provided Figure 2 in black and white to match the description in the caption.
We have made appropriate changes in the caption description of Figure 4.
Referee: 2
1) In the calculation of the false alarm probability (Appendix A), I
do not see how the derivation takes into account the fact that the
number of candidates N_l per segment is itself a random variable
(presumably having a Poisson distribution). It seems that the final
expression for P_FA depends on the N_l's, but their values depend on
the specific noise realization in the data that was analyzed. Hence,
P_FA also appears to depend on the specific noise realization, which
does not make sense to me. The authors should clarify this point and,
if needed, elaborate on it in the text of the paper.
R: For the search that we have performed we obtained a specific number
of candidates N_l in each segment l and the parameter space searched
was fixed and divided in a specific number of cells.
For a given search the N_l and N_c were fixed and deterministic.
Our formula gives probability of obtaining C_max or more coincidences
by chance for those fixed N_l and N_c values.
For a different search (different data, different search parameters)
the number N_l and N_cell would be different and our formula would give
a different probability.
Our false alarm probability is a probability of a specific occurrence
and is not a random variable in itself.
2) When the summation variable is changed from t to t_b in Eq. 15, and
since the t_b values are regularly spaced, the terms inside the
summation are no longer evaluated at the same values of t as in Eq.
10. (where it was t that was regularly spaced). This means that the
value of each term changes and, hence, the value of the sum. Thus, the
F_a or F_b in Eq. 15 are not the same as in Eq. 10. I suspect the
change in these quantities is not very significant, but the
expressions should be suitably corrected (or the associated text
modified) for the sake of completeness. (Note that Eq. 22 includes
small correction terms like T_obs*v_max/c to T_obs.)
R: This is correct. The values of F_a or F_b in Eq.(15) are not the same as in Eq. 10.
We obtain the values of F_a or F_b at t_b from the values at t by
using an accurate interpolation procedure described in Section VIA of
Astone et al., 2010 (Phys. Rev. D 82 022005).
We have added text below Eq.(15) explaining this.
3) After Eq. 1: "Let N0 be the number of zeros in a given data
segment". Since the data segments are in the time domain, and noise
has mean zero, there must be a large number of samples close to zero.
Is N_0 the number of samples with an *exact* value of zero? If so, it
will benefit readers if this is stated explicitly in the text.
R: The N_0 is the number of samples with their value exactly equal to 0.
A datum is set to 0 whenever it is considered "bad".
These bad data occur whenever the detector is not working properly
and these data are characterized as non-science data.
As they are set to zero manually it does not mean that there are
many data points close to 0.
We have changed the explanation in the text about zero samples to:
"Let $N_0$ be the number of zeros in a given data segment
(a data value at a given time is set exactly
to $0$ whenever at that time there is no science data)."
4) Page 12. What was the reasoning behind the selection criteria (1
and 2) listed at the beginning of this page? Some explanation should
be provided.
R: There is no deep reason behind these criteria.
They are a rule of thumb.
We have changed sentence
"if the following two criteria are met"
to
"if the following two criteria, selected as a rule of thumb, are met"
5) The caption of Fig 2 refers to a gray shaded image while it is in
color in the pdf version. It is not easy to understand which parts are
gray from the color version.
R: We have provided a black and white version of Figure 2 to match
the caption description.
6) Sec 5, 1st para: "maximizing the likelihood function" should
probably be "maximizing the likelihood ratio function" (even though
they are mathematically identical when the noise parameters are fixed
in the maximization).
R: Yes, we agree. We have changed the text accordingly.
7) Sec 9, 1st para: Given the very large numbers of templates,
candidates etc., quoted in this paragraph, it will be interesting to
learn what computing resources were used for the analysis. I suggest
adding a brief description to that effect.
R: We have added two sentences after the 1st sentence of Sect. 9.
8) In Fig. 7, the subpanels showing the time series and the histogram
are missing units/scale on the y-axis.
R: We have corrected Figure 7.
In the new version we have given time series scale in absolute values.
The amplitude is dimensionless.
The histogram is normalized in such a way that the y-axis gives probability
that the 2 x F-statistic value is within a given bin.
Sum of the probabilities corresponding to the bins is equal to 1.
Probability is dimensionless.
9) Caption of Fig 7.: "The distribution shows singularity at the
ecliptic." What is the nature of the "singularity" that is being
referred to? The origin of this gap in the distribution should be
elaborated upon in the main text.
R: We have added text explaining this immediately after Figure 7.
============================================================================