Filter tailing at gaps and extreme ends of data - how to avoid?

I am using a Butterworth filter to remove the periodic signal from a data series. However at the ends (and at gaps) of the time-series the filter tails off? I attach an image to describe the effect I am talking about. The red line shows the filtered data, note that at the end of the series when the unfiltered data rises dramatically, the filter does not capture this effect.

Why does this happen and is there anyway of avoiding it?

This question is also linked to the following question: http://www.mathworks.com/matlabcentral/answers/159813-filling-data-gaps-in-a-periodic-timeseries-in-matlab

Answer 1

[Image Analyst]

Try copying the last elements (an amount equal to the filter window width) and then filter the extended signal. Then crop off the copied elements to get back to the original size. Not perfect but should be an improvement I think.