Using the IGS orbit accuracy codes 

To obtain the best results with the IGS products the user is advised
to use the accuracy codes provided in the header of the orbit (sp3) files.
According to the SP3 documentation, an accuracy code N denotes a position 
accuracy (one sigma) of +/-2 mm to the power of N. The quoted orbital error
(1-dimensional) represents an average over the whole period covered by the
SP3 file, based on the mutual RMS differences among the contributed orbit
solutions used in the IGS combination (after adjusting a 7-parameter Helmert
similarity transformation for each solution to account for possible systematic
frame differences). These accuracy codes should be used to differentially
weight the satellites or, in case the user software does not provide the
possibility to individually weight the different satellites, the accuracy
codes may be used to remove "bad" satellites (e.g., accuracy code > 6 for
Final products).

Some satellites have an accuracy code of zero (0). These are satellites
for which only one of the IGS Analysis Centers provided a solution. These
satellites should probably be discarded in the GPS data processing or at
least they should be heavily down-weighted. Most likely these satellites
suffered from significant modeling problems for the day in question and/or
were undergoing orbit maneuver events.

Note that starting in August 2004 the IGS combined orbit products began to
include similar time-dependent accuracy estimates for each satellite and
each 15-minute epoch in the extended sp3c format. These time-dependent
error estimates are not very different from the aggragate header accuracy
codes except when Ultra-rapid predictions are included. Note also that
using mutual agreement among the available orbit solutions as the accuracy
metric almost certainly leads to overly optimistic assessments. These
should, however, still be useful for flagging orbits of dubious quality.