Switch Null Hypothesis

[mwilensky768]

The current match filter asks if data over a specific range of frequencies match the expected noise level. If the data exceed the expected noise level, data in that range are flagged as RFI. In other words, We ask if the data match noise, not if the data match an expected RFI shape.

A match shape filter would proceed in the following way:

(1) Input a shape over a given frequency range, rather than just a frequency range.
(2) Calculate the residuals between the shape and the data.
(3) If the sum of the residuals are beneath a certain threshold, these are the data you're looking for.
(4) If the residuals are above a certain threshold, these are not the data you're looking for.

I think this will catch less RFI, but it will prevent the problem of flagging streaks every time there is a shape in the data which is not accounted for in the filter settings.