This page presents the first tests of an Epoch 4 background model constructed from ~3/4 of the first year of Epoch 4 blank sky data. The resulting models are tested against both the data that went into the model and the data that was reserved. The models are pca_bkgd_faintl7_e4v20000625.mdl and pca_bkgd_faint240_e4v20000718.mdl .
Rick Edelson and colleagues suggest a method of determining the error
associated with each point as follows. Estimate the background corrected
rate for each 16 sec interval; for each series of contiguous 16 second intervals,
estimate the mean, standard deviation, and standard deviation of the mean.
I have done this for data from each of 5 blank sky fields, selecting only data
with 30 or more consecutive 16 second intervals (comparable to the shortest
RXTE observations). The division of "model" and "test" data ensures that the
times are interleaved (although this is not shown in the presentation below).
For each pointing direction, I collected the channel 0-23 (~2-10 keV) light
curve, taking data from 2 detectors (PCU 0 and 2). I calculated the quantities
above, and plots standard-deviation of the mean vs number of intervals.
The black points are from data that went into the model; red points are
from data that did not. The mean rate may be different in the different blank fields,
but this presentation supresses that difference. The assymptotic value of
of the standard-deviation of the mean can be thought of as an estimate of
the short term systematic error. (Estimates previously presented by KJ and
DAS are primarily long term; similar estimates will be produced for this
data set shortly.)
Point 1, layer 1, 2-10 keV errors on the mean
Point 2, layer 1, 2-10 keV errors on the mean
Point 3, layer 1, 2-10 keV errors on the mean
Point 4, layer 1, 2-10 keV errors on the mean
Point 5, layer 1, 2-10 keV errors on the mean