Page 1 of 1


Posted: Tue Oct 04, 2005 7:14 pm
by cringle
When applying FIMIX-PLS, Hahn et al. (2002) suggest using only positive values – if theoretically reasonable – for the relationships in the structural model. Is it possible to implement such restrictions into SmartPLS 2 in order to restrict the weights in the inner model as well as the variance of the latent endogenous variables for the FIMIX-PLS procedure?

Hahn, C., Johnson, M. D., Herrmann, A. and Huber, F. (2002) Schmalenbach Business Review, 54, 243-269.

Posted: Thu Oct 06, 2005 2:13 pm
by awill
Such restrictions of the segment specific relationships in the structural model might help to receive an appropriate FIMIX-PLS segmentation. I think it's possible to implement this capability in the next SmartPLS version.

t-values in FIMIX-PLS

Posted: Fri Nov 18, 2005 8:35 am
by thorsten.giesa
great work on the FIMIX-PLS program! However, I have one suggestion for a future release of smartPLS. Would it be possible to add the t-values for the path coefficients of the respective segment solutions, so one can make statements about their significance? This feature is implemented in the standard PLS routine in smartPLS and would be of great in help in interpreting the FIMIX-PLS solutions!
Thank you and keep up the good work!

t-values in FIMIX-PLS

Posted: Tue Dec 20, 2005 7:58 am
by thorsten.giesa
Concerning the t-values in the FIMIX-PLS option again: Is it planned to implement this feature in the next release of smartPLS? In the meantime, I have seen that in your article: FIMIX-PLS: Methodology and Application ( on page 5 first paragraph that you have calculated t-values for the FIMIX-PLS solution: "...analysis of FIMIX-PLS results does not give evidence for the relevance of this variable (t-statistic=0,690)". I haven't found standard deviations for the FIMIX-PLS Path Coefficients in the output, is there any other way that I can calculate them myself with the given information in the smartPLS output? Any help would be greatly appreciated!

Posted: Tue Dec 20, 2005 10:44 am
by cringle

these t-statistics refer to the posthoc analysis by Ramaswamy et al. as proposed by Hahn et al. (viewtopic.php?t=21), pp. 262-263, and not to the FIMIX-PLS procedure. This posthoc analysis in not implemented into SmartPLS (and probably will not be because concerns exist that this method does not reliably identify explanatory variables for segmentation).


Posted: Tue Dec 20, 2005 10:59 am
by thorsten.giesa
Thank you for clarifying this point! As the pages prior to the one mentioned in my post before are missing from the downloadable pdf, I just assumed the t-values mentioned were regarding the FIMIX-PLS path coefficients. Nonetheless, while the posthoc analysis will not be implemented, what about the t-values concerning the path coefficients in the FIMIX-PLS Procedure? Hahn et. al (2002) clearly used those in their article on page 261, Table 5. Thanks again for clarifying!