Different softwares, different results

Log Out | Topics | Search Moderators | Register | Edit Profile

Author Message   Hongtu Xie (tonali) Junior Member Username: tonali Post Number: 6 Registered: 2-2010 Posted on Monday, March 08, 2010 - 8:08 am:    ok, Thanks Michel, I'll check them according to your suggestions Many thanks!   Michel Coene (michel) Senior Member Username: michel Post Number: 48 Registered: 2-2002 Posted on Monday, March 08, 2010 - 5:03 am:    + Make sure the two packages do their estimation for the same number of PC's. + Small differences can occur depending on how the sample set is divided in calibration / validation + With a cross-validation there can be a randomisation effect in the choice of segments. + You could introduce an error while importing/exporting data, such as comma/point as decimal separator in ascii files.   Hongtu Xie (tonali) New member Username: tonali Post Number: 5 Registered: 2-2010 Posted on Thursday, February 25, 2010 - 10:48 am:    I got different resuls from different softwares, i.e, I use OPUS6.5 and PLS-toolbox5.5 for calibration. The R^2, RPD in PLS-toolbox5.5 were lower than OPUS6.5 using the same training set. What's the reason? Maybe the algorithm is different between the two softwares? I'll test them in detail.