Selecting Moderator Variable

fairviewpark · Post by **fairviewpark** » Tue Apr 08, 2008 10:42 am

Hi PLS experts,

I tried assessing the moderating effect with SmartPLS. When it comes to selecting a moderator, I was stuck because my moderator was not there for selection. I believe the reason was that the moderator was not incorporated into the model. The moderator is "purely" a moderator and so does not form an endogenous or exogeneous construct in the "base" model. How may I get around it? Your help is much appreciated.

Fairview

Diogenes · Post by **Diogenes** » Thu Apr 10, 2008 7:54 pm

Hi Fairview,
the way of modeling is like in the regression analysis, first we try just with the Independent variable and the moderator and second we include the multiplicative term (interaction).
As you don´t want to include the moderator as a exogenous variable, the SmartPLS couldn´t compute the multiplicative terms.
To do what you want, you should compute the multiplicative terms before import the data to SmartPLS.
Best regards.
Bido

fairviewpark · Post by **fairviewpark** » Fri Apr 11, 2008 2:30 am

Hi Prof. Bido,

Thank you very much! Would you please advise how to compute the multiplicative terms? I found a post by you on that but still am not sure about how to do that. Would you share or direct me to a resource for the details on how to do it, say in SPSS or Excel? Many thanks!

Best regards,
Fairview

Diogenes · Post by **Diogenes** » Fri Apr 11, 2008 2:47 am

Hi Fairview,
1) All LVs are reflective, for instance:
exogenous) LV1 -> X1, X2, X3
moderator) LV2 -> X4, X5, X6
interaction) LV -> X1*X4, X1*X5, X1*X6, X2*X4, X2*X5, X2*X6, X3*X4, X3*X5, X3*X6
These multiplicative terms are computed in Excel or SPSS and are imported to SmartPLS to be used as indicators of the interaction LV.

2) some of the LV is formative:
- run a model with this LV as exogenous to have its scores computed
- copy/paste to Excel or SPSS
- compute the single indicator score_LV1 * score_LV2
- this single indicator will be used as the indicator of the moderator.

Best regards.
Bido

hongling · Post by **hongling** » Sat Nov 08, 2008 11:53 am

Dear Prof. Bido,

for the second case (some of the LV is formative), should we standadize the scores of the LV1 and/or LV2 before computing the product LV1* LV2? Or should we standadize the product after it is made?

Or perhaps there is no need at all to consider the standadization issue?

Thanks for your reply in advance!

Best regards,
Hongling

iris_afandiphd · Post by **iris_afandiphd** » Mon Jul 05, 2010 2:01 am

Dear Prof. Bido,

Do you have picture about the process?
Thanks