contradictory regression/model results

Research topics can be discussed in this area.
Post Reply
Fabian S.
PLS Junior User
Posts: 1
Joined: Wed Jan 27, 2016 12:27 am
Real name and title: Fabian Schüssler

contradictory regression/model results

Post by Fabian S. »

Hey guys,
I am not a native English speaker, if anything is unclear, just ask me.

In a survey with N=127 participants every participant got randomly assigned two variables (in the pictures: "Zugewiesene Variable 1"/Severity of Penality and "Zugewiesene Variable 2"/ Certainty of Detection).
The first assigned variable (Severity) has the values 1, 2 and 3. The second assigned variable has the values 0 and 1.

Now in the first picture (Small Model) you can see the default bootstrapping on the variable on the right ("Untersuchende Variable"/ Intention to Comply), which the users rated. In the brackets you can see the p-values are not significant.

This seemed strange to me, since I knew that other variable called Abbildende Variable 2 had a significant positive effect on Intention to Comply/ "Untersuchende Variable". "Abbildene Variable 2" and "1" have the same meaning like Severity and Certainty, but are not assigned to participants, but rated by them.
This meant for me, there must be a non significant effect between the assigned variable and its by the participants rated variable. Otherwise there would be an indirect significant relationship, and that surely must mean, that there has to be a direct significant relationship!?
Another thing to notice is that the R^2 is near zero.

That was the reason I made a second model called big model, which you can see in the second picture.

The second picture strangely has one relationship, that now is significant. And it is negative significant. Another miracle for me is, that the indirect relationship is positive and significant, but the direct relationship is significant negative. (I am talking about the lower triangle).

Small:
Image
Big:
Image

If you can help me out with these questions and point out my mistakes, I would be very thankful.

I could also upload another picture, where Abbildende V 1 is not significant to Untersuchende V and Abildende V 2 is significant to Untersuchende V.
User avatar
Hengkov
PLS Super-Expert
Posts: 1599
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Re: contradictory regression/model results

Post by Hengkov »

Hi,

1. For model 1, you use the ordinal variable, either in independent or dependent, it may not be appropriate and take appropriate algorithm, so make your model badly.
2. You have a problem supression effects in model 2, that's it.

Best regards,
Post Reply