Very large AR regressions

Very large autoregressions

Heavily optimized for very large autoregressions (ie. 100 millions observations and 500 lags). Doesn't consume additional memory. Up to x100 increase in speed then running regressions on very large datasets vs naive estimation.

example:
[xtx, xty] = xL(y(2:end), [y(1:end-1) y(1:end-1).^2], [200 100]);
beta = inv(xtx)*xty;

runs regression of y on 200 lags of y and 100 lags of y.^2