khliland / plsvarsel Goto Github PK

Hi,

I was testing the function WVC_pls and the weighting factor in code is equal to : $\alpha_a = \frac{w_a^T X_a^T Y_a q_a}{w_a^T X_a^T X_a w_a}$. However, in the research article it says that it uses $\alpha_a = \frac{w_a^T X_a^T Y_a q_a}{w_a^T X_a^T Y_a w_a}$.

I would like to know which of the two weighting factor was used for simulation in that article.

Thank you!

Hi Kristian. A suggestion:
##plsVarSel::sMC function function (pls.object, opt.comp, X, alpha_mc = 0.05) { b <- pls.object$coefficients[, 1, opt.comp] X <- unclass(as.matrix(X)) n <- dim(X)[1] yhat <- X %*% b Xhat <- tcrossprod(yhat, b)/crossprod(b)[1] Xresidual <- X - Xhat SSCregression <- colSums(Xhat^2) SSResidual <- colSums(Xresidual^2) MSCregression <- SSCregression MSResidual <- SSResidual/(n - 2) smcF <- MSCregression/MSResidual smcFcrit <- qf(1 - alpha_mc, 1, n - 2) attr(smcF, "quantile") <- smcFcrit smcF }
I reckon it will not be clear to many users that the sMC function returns the F-values (of each variable) and the critical (threshold F-value).
In other words, most will assume that the value returned is the p-value of the F-test.
It would be worthwhile to add to the documentation.