Incorrect Wild bootstrap overall

[FinYang]

Wild bootstrap should have $\hat\epsilon_t\nu_t$ where $\hat\epsilon$ is the residuals and $\nu_t$ is randomly simulated from the chosen distribution. Right now the implementation in the package all uses $\hat\epsilon_t + \nu_t$. For example,

tsDyn/tsDyn/R/VAR.sim.R

Lines 190 to 194 in 15c8679

	innov <- switch(boot.scheme,
	"resample"= resids[sample(seq_len(t), replace=TRUE),],
	"wild1"=resids+rnorm(t),
	"wild2"=resids+sample(c(-1,1), size = t, replace=TRUE),
	"check"= resids)

[MatthieuStigler]

thanks, I think you spotted a mistake indeed. It has been a while so I can't remember if I had followed a paper suggesting this, or whether that was a blunt mistake of mine, but looking at a few papers right now, it seems indeed multiplying as you suggest is the right approach.

Thanks again!