GALEX photometry tests (July 2008)
Approach
Method
Results - Mann Whitney
Results - 3D Correlation Functions
Results - Angular Correlation Functions
Approach
The Angular Correlation Function (ACF) derived from GALEX fields shows
peculiar behaviour that could be due to effects similar to
field-to-field fluctuations. In order to investigate the origins of
these effects, we performed several tests, based mainly on the
Mann-Whitney statistic (a rank test), that we use to quantify the
"quality" of the photometry of a given field compared to the rest of
the fields. This tests showed that the ACF is very sensitive to a
field selection by this quality parameterization, and that the
Mann-Whitney statistic is effective to pick up potentially problematic
fields (the ACF of worse quality fields according to the Mann-Whitney
statistic show stronger effects similar to field-to-field fluctuation;
see the
previous tests ).
A solution to get rid of these effects while measuring the ACF is to
perform the measure independently on each field. This method can not
however be used for the measure of the 3D correlation function using
SDSS spectroscopic redshifts, as the statistics are fairly low.
Hence we performed additional tests on GALEX pipeline photometry, in
order to understand whether the 3D clustering measurements are
affected in the same way than the angular ones.
Method
To derive the Mann-Whitney test, we consider GR3 AIS or MIS fields, using:
- FUV or NUV exposure times greater than 1000s. (for MIS)
- Objects within the primary resolution and with fov_radius < 0.5 deg
- Objects with UV magnitudes: 17 < UV < 22 (or 20.5)
- Objects with one SDSS counterpart within 4 arcsec
- Objects with SDSS galaxies (type = 3) counterparts, to avoid any
effects due to variation of stars counts with Galactic latitude.
We run the Mann-Whitney test on the UV magnitudes of the objects of
each field considered as a test sample, while the remaining objects
that belong to other fields build the control sample.We use, as output
of the Mann Whitney test, the p-value, that provides the probability
that the test and control samples are drawn from the same
distribution. We consider a threshold in p-value (< 0.05), to
classify the fields as outliers in terms of photometry.
We focus in particular on AIS fields overlapping MIS fields.
Results - Mann Whitney
These first plots show the histogram of the p-values of the fields;
here we consider only overlapping AIS and MIS fields. The left plot
shows AIS selection with UV< 20.5 and MIS UV< 22, while the right
one shows results from same cuts in both surveys (UV< 20.5). AIS
(MIS) histograms are plotted in solid (dashed )lines; FUV is blue, and
NUV red. Each plot also shows the percentage of fields with a p-value
lower than 0.05 (potentially outliers).
The fraction of outlier fields is larger according to NUV magnitudes,
and also when including fainter objects.
We look here only at MIS fields, but considering now co-adds and
non-coadds. The left plot shows, as, before, the histograms of the
p-values for co-adds (solid lines) and non-coadd (dashed lines). As
before, FUV is blue, and NUV is red. The percentage of outlier fields
is larger within coadds. The plot on the right hand side shows these
percentages.
At brighter magnitudes (UV < 20.5), the percentage of outlier fields
is in agreement between coadds and non-coadds.
The two plots above show the percentage of outlier fields as a
function of the number of coadds; UV < 22 on the left, and UV <
20.5 on the right. The histograms (using the left axis) show the
number of fields in each bin. These plots suggest there is a trend
(for UV < 22) that fields with a larger number of coadds are more
likely to be outliers according to the Mann Whitney statistic.
Results - 3D Correlation Functions
We present here some results of 3D correlation function (actually
projected correlation function). We use here GALEX counterparts of
SDSS spectroscopic galaxies.
AIS fields selected by Mann-Whitney
The plot on the left shows the projected correlation function for
NUV< 19, from AIS fields. Black circles represent the correlation
function of the full sample; blue triangles the correlation function
of "good quality fields" according to Mann-Whitney, and red squares
outlier fields. The outlier fields have a very different correlation
function, with shallower slope, and higher amplitude. The lines
represent best fits by power laws. Fainter samples or FUV selections
show similar trends.
The right handside plot shows the chi2 contours in the
(slope, correlation lengh) plane of the correlation function power law
parameterization. The outlier fields sample has a correlation length 2
times larger than the other fields. It should be noticed that this cut
(NUV< 19) ensures a UV-selected sample, and also contains almost no
galaxies on the red sequence. This clustering strength is similar to
what is expected for fairly large galaxy groups. Hence, this strong
clustering can be fake (due to some photometry problems) or actually
real, if these fields contain a significant number of groups or
clusters. It should also be noticed that, contrary to angular
correlation functions, these projected spatial correlation functions
look "reasonable".
MIS fields coadd vs non
coadd
These plots show the projected correlation functions for FUV < 22
(left) and NUV < 22 (right) from the MIS fields with coadd (red) or
non coadd (blue). Although there are some differences between coadd
and non coadd fields according to the Mann-Whiney test, there is no
obvious trend when considering the correlation functions.
SDSS data within AIS fields selected by Mann-Whitney
Here we used SDSS data within AIS fields; the correlation function has
been computed for spectroscopic galaxies with r < 17.6. The only
GALEX information used here is the result from the Mann-Whitney test
(NUV magnitudes in this case). As before, there is a clear trend that
fields with low p-values from the Mann-Whitney statistic based on UV
data have a stronger clustering. This suggests that the effect
observed for the projected correlation function from GALEX UV-selected
data is real.
Mann Whitney and clusters
?
The previous results suggest that the Mann-Whitney statistic could be
sensitive to galaxy groups or clusters. This plot shows one result
obtained from the application of the friend of friend algorithm on the
SDSS spectroscopic galaxies (by M. A. Aragon-Calvo). Here we use the
AIS fields, and plot the ratio (normalized by the number of fields) of
the number of "outliers" fields to the number of "good quality" fields
as a function of number "cluster" members. There is a slight trend, in
NUV, that "outliers" fields contain more numerous clusters.
Results - Angular Correlation Functions
The global trends of the angular correlation function with the results
from the Mann-Whitney statistic have been presented elsewhere
. Some new test are showed here.
MIS fields coadd vs non
coadd
The angular correlation functions of MIS fields coadd (filled squares)
and non coadd (filled circles) are presented here, for various cuts in
FUV (left) or NUV (right). As before with the projected spatial
correlation function, there is no obvious trend between coadd and non
coadd fields in terms of measured clustering. Note however the
flattening of the angular correlation function at scales larger than
0.05 degrees.
SDSS data within MIS fields selected by Mann-Whitney
These angular correlation functions have been computed using SDSS r
selected data (r < 19) within MIS fields. As before, the fields have
been selected according to the p-values from the Mann-Whitney test,
ran on FUV magnitudes (left) or NUV magnitudes (right). The black
squares show the angular correlation function of the full sample,
while the blue circles show the result from the fields with "good
quality" in terms of Mann Whitney statistic, and red triangles results
from the "outliers". Note here that the correlation functions do not
show flattening as important as the one seen with GALEX data (the SDSS
angular correlation functions are well approximated by power laws from
0.01 degrees to large scales). This suggests that there are some
field-to-field variations within the GALEX photometry that affect
mostly angular correlation functions.
