Interaction Effect: IV formativ, MV reflectiv, DV reflectiv

Romschti · Post by **Romschti** » Sat Feb 04, 2012 5:34 pm

Dear Researchers,

there is (simplified) a structural model:

IV => DV.

The IV is formativ, the DV is reflectiv and the MV is reflectiv (metric). What`s the right procedure to test moderation?

1.) Two-Step-Approach (latent variables scores of DV and MV, create Interaction Variable in Excel, ...)

2.) Implementation of the MV with SmartPLS function "Create Moderating Effect"

3.) Any other possibilities (no multi group)?

Thank you so much!

Hengkov · Post by **Hengkov** » Wed Feb 08, 2012 1:09 pm

Hi Kathirn,
Two-stage approach appropriate procedure for test your model interaction.
Stage 1: the main effect PLS model is run in order to obtain estimates for the latent variable scores. The latent variable scores are caculated and saved for stage 2 analysis.
Stage 2: the interaction term X x M is built. This interaction term as well as latent variabel scores of X and M are used as IV in a multiple regression on the latent variabel scores of Y.
Regards,
Hengky

Romschti · Post by **Romschti** » Sat Feb 11, 2012 8:00 am

Thank You!

Romschti · Post by **Romschti** » Sat Feb 11, 2012 4:45 pm

Excuse me. I`ve got one more question:

Should I use the

a.) standardized latent variable scores (LVS) or the
b.) unstandardized latent variable scores (LVS)

to calculate the moderating effect ("two-stage-approach")?

Thank you.

Hengkov · Post by **Hengkov** » Sat Feb 11, 2012 5:12 pm

Hello Kathrin,
Your welcome and you can used standardized latent variabel score. Good luck
Regards,
Hengky

Romschti · Post by **Romschti** » Sun Feb 12, 2012 8:02 am

Great. Thank You.

imran77 · Post by **imran77** » Tue Apr 03, 2012 4:41 am

The IV is reflectiv the DV is formativ, and the MV is reflectiv.

What`s the right procedure to test moderation? two step approach?

How is it being done?

-imran-

Hengkov · Post by **Hengkov** » Tue Apr 03, 2012 8:11 am

Hi Imran,
For your model above, Product Indicator Approach appropriate procedure for test the interaction, because your predictor and moderator variable is reflective.
Your create interaction terms => indicators IV*MV.

IV => DV
MV => DV
Interaction terms => DV

Run algorithm PLS => evaluation outer model and Bootstrap procedure => evaluation inner => done.

For addition => asses effect size interaction= R2 with moderator - R2 without moderator : 1-R2 with moderator.
Regards,
Hengky

imran77 · Post by **imran77** » Wed Apr 04, 2012 6:50 am

I don't get the formula:=

effect size interaction= R2 with moderator - R2 without moderator : 1-R2 with moderator. -- what does this mean?

kristianak · Post by **kristianak** » Wed Apr 04, 2012 8:45 am

Imran,

You should test your model with and without the moderator.

Cohen's effect size - f2 - is calculated as follows:
f2 = (R2_included - R2_excluded) / (1-R2_included)

As Henseler, Ringle and Sinkovics (2009) writes, Cohen's effect sixe is calculated as the increase in R2 relative to the proportion of variance of the endogenous latent variable that remains unexplained. Effect size values of 0.02, 0.15, and 0.35 can be viewed as a gauge for whether a predictor latent variable has a weak, medium, or large effect at the structural level.

HENSELER, J.; RINGLE, C. M.; SINKOVICS, R. R. The use of partial least squares path modeling in International Marketing. Advances in International Marketing, v.20, p.277-319, 2009. Available at: <http://php.portals.mbs.ac.uk/Portals/49 ... cs-PLS.pdf>. - especially pp 303-04

Kristian

Hengkov · Post by **Hengkov** » Wed Apr 04, 2012 11:55 pm

Hi Imran,
You test main effect model without moderator and with moderator => look R-squares two model and calculate effect size interaction with formula above =>
for this case you can used function redo/undo SmartPLS.
Effect size 0.02, 0.15 and 0.35 => weak, moderate and strong.

Hi Kristian,
You right ;-)

Regards,
Hengky