How to calculate filter window width

Hi,

My task is to detrend raw phase and amplitude recorded by the GPS receiver. In the creation of detrending filters, I used article:

Van Dierendonck, A. J., John Klobuchar, and Quyen Hua. "Ionospheric scintillation monitoring using commercial single frequency C/A code receivers." proceedings of ION GPS. Vol. 93. 1993.

I would like to use whole signal recorded during one hour of measurement, instead of discarding first and last 3-4 minutes of the signal due to the edge effects. Possible solution is data expansion by regression or symetric reflection by "half the filter window width".

Problems:

1. How to calculate "half the filter window width"? I am using three 2nd order Butterworth filters (hugh pass for phase, low pass for amplitude), as stated in the article mentioned above.
2. Why is edge effect more pronounced for phase than for the amplitude of signal?

Can anyone suggest any literature on data expansion?