I can uniformly randomly shuffle a collumn and have it as an input!!
That is a "random evil twin". It is a very strong proposition for eliminating false positives in variable importance screening.
Links:
It is huge. I want to do it. ... in JMP.
How do I enter:
- 20 copies of my input column
- and 20 randomly permutated copies of my input colum
- and 20 uniformly randomly generated input columns
as response variables for predictor screening using JSL?