Hello everyone,
the sources of package{plspm} give me some headaches.
rounding:
Firstly, rounding is used on different occasions where I would
not suspect it. For an example, please have a look at the sources of
plsreg1{plspm}. I wonder why to perform rounding within the
function, and not when presenting the results. I would not assume it
would be a performance issue to do that. The rounding is not mentioned
in the documentation either.
Q squared:
Another detail that worries me is that component selection in PLS
regression, i.e. calculation of Q2, seems to me as not correctly
implemented on the first (and second) thought. Again, please see the
aforementioned function for an example; calculated as
Q2[h] = 1 - PRESS[h]/RSS[h]
while it should be calculated as
Q2[h] = 1-PRESS[h]/RSS[h-1]
according to Wold et al 2001, Chemometrics and Intelligent Laboratory
Systems 58(2): 109-130 (among others). Does anyone have some ideas why
it's coded that way? Am I missing something?
Disclaimer: I tried to contact the maintainers, but I recieved no
answers except an autoresponder on behalf of Gaston Sanchez. (I also
tried some other of his e-mail addresses I found on the web, no luck.) I
assume the maintainers are still somewhere around, even if my mails
remained unresponded. The package did not turn up in the orphaned
package list.
So, anyone who could help me on this?
Regards,
arne
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
