Dear list,
I have a question concerning the above mentioned methods in the pls
package with respect to the loadings matrix produced by the call. In
some work I am doing I have found that the values produced are nearly of
the same magnitude but of opposite sign. When I use the example data
(sensory) I find this result reproduced. I am prepared to work this
through but I have a feeling that there could be a possible error in the
code. (?!)
> sens.pcr$loadings
Comp 1 Comp 2 Comp 3 Comp 4
yellow 75.186621 -0.4780473 3.212149 1.750123
green -90.490256 8.5880530 1.634961 1.042239
brown -2.861241 -11.3600509 -15.920789 -1.105799
glossy 13.347090 19.3103902 -3.121693 2.781282
transp 20.126987 24.0653312 -6.656764 -1.842907
syrup -7.199972 -5.3436196 -5.073675 5.620454
> sens.pls$loadings
Loadings:
Comp 1 Comp 2 Comp 3 Comp 4
yellow -74.448 -10.519 3.169 -1.056
green 88.299 21.627 -0.521 -0.976
brown 4.959 -14.253 -12.761 -0.371
glossy -15.798 15.914 -7.574 3.504
transp -23.049 18.673 -12.214 -2.068
syrup 8.045 -5.313 -3.698 2.181
Thank you for your help.
Roy Little
Dept. Chem.
Facultad de Ciencias
Universidad de los Andes
M??rida, Venezuela
Prof Brian Ripley
2005-Nov-22 17:41 UTC
[R] loadings matrices in plsr vs pcr in pls pacakage
Please read ?princomp to understand more about loadings and their arbitrary signs. On Tue, 22 Nov 2005, Roy Little wrote:> Dear list, > I have a question concerning the above mentioned methods in the pls > package with respect to the loadings matrix produced by the call. In > some work I am doing I have found that the values produced are nearly of > the same magnitude but of opposite sign. When I use the example data > (sensory) I find this result reproduced. I am prepared to work this > through but I have a feeling that there could be a possible error in the > code. (?!) > > > > > sens.pcr$loadings > Comp 1 Comp 2 Comp 3 Comp 4 > yellow 75.186621 -0.4780473 3.212149 1.750123 > green -90.490256 8.5880530 1.634961 1.042239 > brown -2.861241 -11.3600509 -15.920789 -1.105799 > glossy 13.347090 19.3103902 -3.121693 2.781282 > transp 20.126987 24.0653312 -6.656764 -1.842907 > syrup -7.199972 -5.3436196 -5.073675 5.620454 > > sens.pls$loadings > > Loadings: > Comp 1 Comp 2 Comp 3 Comp 4 > yellow -74.448 -10.519 3.169 -1.056 > green 88.299 21.627 -0.521 -0.976 > brown 4.959 -14.253 -12.761 -0.371 > glossy -15.798 15.914 -7.574 3.504 > transp -23.049 18.673 -12.214 -2.068 > syrup 8.045 -5.313 -3.698 2.181 > > Thank you for your help.-- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595
On Tue, 2005-11-22 at 08:53 -0400, Roy Little wrote:> Dear list, > I have a question concerning the above mentioned methods in the pls > package with respect to the loadings matrix produced by the call. In > some work I am doing I have found that the values produced are nearly of > the same magnitude but of opposite sign. When I use the example data > (sensory) I find this result reproduced. I am prepared to work this > through but I have a feeling that there could be a possible error in the > code. (?!)But only for the first axis! I am not an expert, but as I understand it, PCR and PLS are related but different methods. PCR components maximise variance within the predictor variables, whilst PLS components maximise the covariance with the response variable. Both methods appear to extract first components that are similar - perhaps indicative of strong structure within the data set. That the subsequent components are different reflects the different criteria maximised by the two methods. Prof. Ripley has indicated where to look for discussions about the signs of eigenvectors. Maybe I misunderstood your query? Were you expecting the results to be the same on all components? Why do you suspect an error in the code? HTH, Gav> > > > sens.pcr$loadings > Comp 1 Comp 2 Comp 3 Comp 4 > yellow 75.186621 -0.4780473 3.212149 1.750123 > green -90.490256 8.5880530 1.634961 1.042239 > brown -2.861241 -11.3600509 -15.920789 -1.105799 > glossy 13.347090 19.3103902 -3.121693 2.781282 > transp 20.126987 24.0653312 -6.656764 -1.842907 > syrup -7.199972 -5.3436196 -5.073675 5.620454 > > sens.pls$loadings > > Loadings: > Comp 1 Comp 2 Comp 3 Comp 4 > yellow -74.448 -10.519 3.169 -1.056 > green 88.299 21.627 -0.521 -0.976 > brown 4.959 -14.253 -12.761 -0.371 > glossy -15.798 15.914 -7.574 3.504 > transp -23.049 18.673 -12.214 -2.068 > syrup 8.045 -5.313 -3.698 2.181 > > Thank you for your help. > > Roy Little > Dept. Chem. > Facultad de Ciencias > Universidad de los Andes > M伱仼rida, Venezuela > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html-- %~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~% Gavin Simpson [T] +44 (0)20 7679 5522 ENSIS Research Fellow [F] +44 (0)20 7679 7565 ENSIS Ltd. & ECRC [E] gavin.simpsonATNOSPAMucl.ac.uk UCL Department of Geography [W] http://www.ucl.ac.uk/~ucfagls/cv/ 26 Bedford Way [W] http://www.ucl.ac.uk/~ucfagls/ London. WC1H 0AP. %~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%