Why PCR function does not give the standard errors of regression coefficients directly? And why are they different from NCSS?

[nursuznb]

I have some problems with PCR function in the PLS package. After doing PCR, I need the original coefficients and their standard errors. I can find coefficients, but I do not get the standard errors directly. I can find them with Jackknife function. But why R gives only Jackknife standard errors? Is not possible to calculate normal variance in PCR? When I do the PCR with the same data in the NCSS program, it gives different coefficients and standard errors to me. But they are always smaller than the original OLS standard errors. But in R, generally, the Jackknife standard errors are bigger than the OLS ones which I got. As you know PCR as a biased technique, standard errors should be smaller than original OLS. Can you help me with this issue?

[bhmevik]

While there is theory for how to estimate standard errors for PCR models, there is no such thing (at least without using things like bootstrap or jackknife) for PLSR models (that I am aware of, at least). The focus of the pls package has been implementation of PLSR models, and PCR was just added because it is so similar and it "completes the picture". Therefore we never added calculation of standard errors to PCR. I can keep this issue open as a feature request and see if we can find time to add it.