[R] PLS Q2 value?

Charles Determan Jr deter088 at umn.edu
Thu May 10 18:02:47 CEST 2012


Greetings R users,

My interest in the Q2cum score comes my endeavor to replicate SIMCAP
PLS-DA analysis in R.  I use the exact same dataset.  After doing the
analysis in R, I can get the exact same R2Ycum.  However, the Q2cum is
significantly off.  Adding the Q2cum of the 1st and 2nd component
comes close but that seems unconvincing and I don't understand why the
3rd component Q2cum would be so far off.  Below is the code and
output:

pls.reg.new=plsreg2(newserumvars.sc, newtimematrix, nc=3)
pls.reg.new$expvar  #this provides the exact R2Y I get in SIMCAP
pls.reg.new$Q2cum

The Q2cum values should be close to
component 1 = .2149
component 2 = .4578
component 3 = .6627 #value of concern

All of the R output values are close for the respective class except
for the final Q2cum value:

   Q2cum.B Q2cum.FR8 Q2cum.S45  Q2cum
t1    -0.0546        0.3992        0.2973   0.2140
t2     0.2863        0.3824        0.7039   0.4573
t3     0.5839        0.7430        0.6980   0.3735  #not close
Done.


I have read that the plspm package used LOO cross validation.  SIMCAP
uses the same method, however it leaves '7' out instead.  Perhaps this
might be an issue that results in this difference?

Any insight or assistance would be most appreciated.  I certainly
appreciate any time you take to answer my question.

Regards,

Charles Determan



More information about the R-help mailing list