Two classes of filtering methods were evaluated. Both made decisions about retaining a probe set in the analysis based on the behavior of the probe set on the arrays in at least one treatment group. If the probe set passes the filter criteria in either group, all signal values for that probe set are retained for both groups regardless of the signal value or detection call for any individual array. Note that this retains probe sets Present in one treatment group and Absent in the other; these can be very interesting biologically. When a probe set does not meet the threshold for either group it is rejected, and signal values for all samples are removed for this probe set. The two methods were applied to data generated by MAS5 and RMA separately. For comparison, filtering was also done using the fraction called present for all arrays.
