Still seeing non-stationary processes

[RhysU]

So much for #3 being closed-- still seeing problems on test cases like...

./lorenz --tfinal=1000 --every=1 --seed=1 | cut -f 4 | ./arsel -s | cut -s '-d ' -f 2-

Compare the above results (showing NaN T0) with that of --every=2. The weird thing is that it doesn't seem to be a problem now with the amount of data (try --tfinal=1500 --every=2 which doubles the data from the above case and contains roughly the same number of points as --tfinal=1000 --every=1). The problem genuinely seems to be the high resolution data.

[RhysU]

Review http://www.citeulike.org/user/RhysU/article/11863472 in light of this issue.

[RhysU]

Ditto for http://www.citeulike.org/user/RhysU/article/11863464

[RhysU]

First two refs indicate that instability due to round-off errors in the time recursion may be the culprit. Current best guess for a more-stable, Burg-like algorithm might be found in http://www.citeulike.org/user/RhysU/article/11863500 (and similar works by Strobach) refered to as PORLA4.

[RhysU]

I am inclined to leave this open but punt for now. Ultimately implementing PORLA4 to see if it solves the issue is probably worthwhile.