p75213
Newbie level 3
Hello,
I have implemented single spectrum analysis and independent component analysis against a time series (in this case the euro foreign exchange) as a smoothing indicator. The ssa algorithm I purchased from: Caterpillar-SSA - https://www.gistatgroup.com/cat/programs.html and the ica algorithm is fastICA from it++ open source software. In both cases the window size is set at 90. SSA has 6 principle components and fasICA has 6 independent components.
In both cases the time series is deconstructed into a matrix of delays before either the ssa or ica processing takes place. After processing the smoothed indicator is reconstructed using diagonal averaging. The strange thing is in both cases the indicators are nearly identical (not quite - observe at the very end where they diverge slightly). Should this be the casel?
Also as new bars are added to the series both indicators readjust themselves retrospectively. How can this action be minimised or better still stopped in either indicator.
I have implemented single spectrum analysis and independent component analysis against a time series (in this case the euro foreign exchange) as a smoothing indicator. The ssa algorithm I purchased from: Caterpillar-SSA - https://www.gistatgroup.com/cat/programs.html and the ica algorithm is fastICA from it++ open source software. In both cases the window size is set at 90. SSA has 6 principle components and fasICA has 6 independent components.
In both cases the time series is deconstructed into a matrix of delays before either the ssa or ica processing takes place. After processing the smoothed indicator is reconstructed using diagonal averaging. The strange thing is in both cases the indicators are nearly identical (not quite - observe at the very end where they diverge slightly). Should this be the casel?
Also as new bars are added to the series both indicators readjust themselves retrospectively. How can this action be minimised or better still stopped in either indicator.