I browsed the net and got to the paper by ANANTHRAM SWAMI AND JERRY M. MENDEL, "Closed-Form Recursive Estimation of MA Coefficients Using Autocorrelations and Third-Order Cumulants" is this the same algorithm used by R (cran) software? MATLAB uses this algorithm I suppose.
My problem is that this paper requires the probability distribution function (pdf) of the input be skewed as a necessity and I feel that it is a strong constraint though most real world problems might have this property satisfied.
Could I relax this constraint or replace with a higher order moment other than skewness (or other values from Information theoratic domain) ?. (NOTE: This Problem arises if I know the output alone and not the input ) else it becomes a simple problem of deconvolution.
- Badges
- Science topic