forum.smartpls.com

Hi everyone,

I have a negative path coefficient between LV and DV , where the LV is likelihood to start a business and DV is categorical variable which is the type of business, so the path coefficient must be positive but I got a coefficient of -.359 and p-value of .085 (>.05).

1-How is that possible ? and how can justify it in my thesis?

2- Can SmartPLS deal with final categorical DV (after dummy-codeding it) ? am confused because i've read in the forum some who says it can it limitation and some says it cant

First, categorical variables do not work well as a DV in PLS as it assumes metric variables like in a regression. Hence, it uses your 1,2,3 (or whatever categories you have) as a 1,2,3 in a metric sense. That could already explain the strange result.

Moreover, if a coefficient is not significant (from zero) it is somewhat meaningless to talk about its sign as it is likely that it could be a null effect.