Interaction effects

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

I'm very glad to hear that!
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Here comes an interesting literature for our topic in this post:

Author: Henseler, J.
Title: PLS-MGA: A Non-Parametric Approach to Partial Least Squares-based Multi-Group Analysis
Year: 2012
Pages: 495-501
In: W. Gaul et al. (eds.), Challenges at the Interface of Data Analysis, Computer Scince, and Optimization, Studies in Classification, Data Analysis, and Knowledge Organization, Springer-Verlag Berlin Heidelberg 2012.
Link: http://ebooks.narotama.ac.id/files/Chal ... alysis.pdf
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

---
Last edited by christian.nitzl on Sat Aug 18, 2012 11:46 am, edited 1 time in total.
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Hi everybody,

Here comes a working paper which is based on my Phd. thesis. In this paper I work with multidimensional constructs and also with the group test by Henseler.

Authors = Nitzl, C./Hirsch, B.
Title = Why Do Managers Trust Their Management Accountants?
Working Paper
Year = 2012
Link = http://papers.ssrn.com/sol3/papers.cfm? ... id=2127701

Comments are highly appreciated!

Christian
Lacra
PLS Junior User
Posts: 10
Joined: Fri Sep 28, 2012 3:21 pm
Real name and title: Radomir Lacramioara

Post by Lacra »

Dear Professor Nitzl,

Could you please send me the Excel template for the moderating effect as well?

I have already read the paper written by Markus Eberl in the Handbook of PLS but I do not understand how he calculated the values corresponding to diff. column in the table on page 507. Since he states: "For example, the impact of performance on likeability is significantly larger in the media group than in the politician group (difference=0.440), or the customers selected from the general public (difference=0.525)", I thought that he calculated the difference in the path coefficients but the results that the author reports do not confirm my presumption.

Thank you
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Sure! Please send me a message to my mail address christian.nitzl@unibw.de. Then I can send you the Excel template.

Best
Christian
samshul78
PLS Junior User
Posts: 7
Joined: Tue Jan 22, 2013 4:29 am
Real name and title:

Post by samshul78 »

Hi Prof Nitzl and everyone
I'm quite new with Smart PLS and SEM altogether..i'm having some problem in understanding how to proceed with my present model using Smart PLS.

Basically the paths of my model is like this
a) Consumer Efficacy --> willing to boycott
b) Consumer Animosity --> willing to boycott
c) Consumer Ethnocentrism --> Prior Purchase --> willing to boycott
d) Consumer Ethnocentrism --> Product Judgement ---> willing to boycott
e) Product Judgement --> Prior Purchase

All indicators are measured by Likert type scale 1-7 except for Prior Purchase. Prior Purchase have 2 dichotomous indicators (Yes/No) answering to the questions of i) Do you consume the product within the last 3 months? and ii) Do you consume the products now?

My questions are :
a) Can i treat the indicators of Prior Purchase as they are (being dichotomous), and run PLS algorithm and bootstrapping? is it allowable? or do i need to do something with the data beforehand?

b) Or do i have to use PLS-MGA - of which to my limited understanding is to seek the differences of sub-groups within a population?

Many thanks in advance. Appreciate all replies.
Thanks.
Thanks

Samshul
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Hi,

at question a) In the actual SmartPLS version it is not possible using a dichotomous variable as dependent variable. For that case you need a logistic-regression. Therefore in your case a valid interpretation of you model is not promising.

at question b) As your question indicates, the question is, if you have theory on hand which allows you performing such a group test? With such a group test you would check if there is a difference between the path coefficients of two separated estimated models.

Greetings,

Christian
samshul78
PLS Junior User
Posts: 7
Joined: Tue Jan 22, 2013 4:29 am
Real name and title:

Post by samshul78 »

thank you for your quick reply, Dr Nitzl.

a) sorry, i am not quite clear on this part -- "In the actual SmartPLS version it is not possible using a dichotomous variable as dependent variable"....does this mean that even if the variable (in the case its Prior Purchase) is both endo and exo in nature, i cannot use it in PLS modeling?

i am quite confused now.

thanks very much.
Thanks

Samshul
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Yes, that does it mean. If a variable is dependent from another variable I think in the actual version of SmartPLS you cannot properly interpret it. Why should be there a different when the dichotomous dependent variable in turn influences other variables?

Greetings,
Christian
samshul78
PLS Junior User
Posts: 7
Joined: Tue Jan 22, 2013 4:29 am
Real name and title:

Post by samshul78 »

Dr Nitzl

OK..thanks..i think i need more readings to understand this area.

regards.
Thanks

Samshul
bwilson
PLS Expert User
Posts: 41
Joined: Wed Apr 02, 2008 7:11 am
Real name and title:
Contact:

Post by bwilson »

Hi,
You should see the PLS Handbook Chapter by Wilson especially for continuous interactions
http://www.springer.com/statistics/comp ... 40-32825-4
Using PLS to Investigate Interaction Effects Between
Higher Order Branding Constructs

Henseler for Multi Groups

regards

Brad
Bradley Wilson. Ph.D.
Senior Lecturer in Advertising.
RMIT University.
School of Media and Communication.
GPO Box 2476V
Location. 9.5.20
Melbourne. Victoria.
Australia.

SEE FOR PUBLICATIONS
www.rmit.edu.au/staff/bradleywilson
malarabi
PLS Junior User
Posts: 1
Joined: Tue Sep 03, 2013 6:36 pm
Real name and title:

Whole Model vs. Specific Relation testing.

Post by malarabi »

Dear All,
The forum has been very helpful.

I have a question about testing for interaction effects:

I have a fairly complex model. I am interested in both explaining the variance for the main dependent variable and testing the effect of a "social" variable(supervisor influence) on most of the model's LVs. Specifically, I am testing for moderation(using a moderating variable) between the relationships between Supervisor Influence and other LVs.

When I run the model as a whole and enter the moderation effects, the model becomes extremely complex and the many of the interaction effects become NS. On the other hand, If I run the interaction on a the specific relation between each variable and the supervisor influence I get Significant results.

Theoretically, it seems fine to test each relation(interaction hypothesis) separately since I am interested in this specific effect.

I have looked for literature on this topic, but there is no mention for the "design" of the run and most papers use the simplest interaction example, which is between two vars and the Moderator.

Advice?

Thanks
saedeh
PLS Junior User
Posts: 2
Joined: Fri Nov 29, 2013 6:14 am
Real name and title:

interaction

Post by saedeh »

Hi
How can I show interaction(refelective relation) in smart pls?(In othet word I want to show A has effect to B and B has effect to A too) but I just found one_way relation

please help me because I need a very fast to your answer
Thanks
bradley7680
PLS Expert User
Posts: 51
Joined: Tue Nov 04, 2014 5:27 am
Real name and title: Dr. Bradley Wilson. Senior Lecturer in Advertising. School of Media and Communication. RMIT University.

Re: Interaction effects

Post by bradley7680 »

Check WILSON contribution on Researchgate

From PLS handbook. Very user friendly

Brad
Bradley Wilson. Ph.D. Senior Lecturer in Advertising. School of Media and Communication. RMIT University. Melbourne, Australia.
SEE: http://www.researchgate.net/profile/Bradley_Wilson2 http://scholar.google.com.au/citations? ... AAAJ&hl=en
Post Reply