non linear relationship and Cox regression

tim.inteeworn · Post by **tim.inteeworn** » Sat Oct 14, 2006 1:30 pm

Hi,

I am not a very experienced user of smart PLS and have the following problem:

I want to model emotions according to the Mehrabian/ Russel approach. My meassurement model are the three dimensions Pleasure, Arousal, and Dominance, which are expected to be formative in nature.

The problem now is, that literature suggests a non linear relationship between arousal and positve emotions, so to speak between a measurement variable and the latent variable. There is indication that there is rather an optimum (arousal was measured on a 5 point likert type of scale, optimum would be 3).

Is it possible to transform this measurement variable with some SPSS procedure with the result of getting a monotone Variable instead of a inverted u shaped variable to use this variable then again according to the linear approach in smart PLS?

As research topic is modelling "time used shopping" and the effects on consumers shopping time, i found cox regression as a useful tool used in statistical analyses. Did anyone use this in addition with PLS? Is it possible to combine the two approaches?

I would be very thankful for pointing me in the right direction or any answers, as I find this quite difficult...

Thanks,

Tim

enijssen · Post by **enijssen** » Mon Apr 28, 2008 7:37 pm

Hi,
Did you ever get any rply to your question. I have modeled some quadratics but get some strange results in one of the pathways. I though standardizing might take care of the problem but note exactely the same results.
So, I started searching on this forum/site. I noted a recent question on quadratics, but no answers yet. I also note you raising a some issue on non-linearity, but see no responses...?!? Any insight you developed since on how to proceed?
Kind regards,

Ed

Diogenes · Post by **Diogenes** » Wed Oct 15, 2008 3:20 pm

Hi Tim and Ed,
see viewtopic.php?p=2415#2415
I hope this help you.
Best regards
Bido