forum.smartpls.com

i have a control variable , which i wan to model in my theoretical framework, for testing, how should i include it, (draw it) , and estimate it,
Control variable here is a latent construct.
But still i am not able to figure out how to include control variables (given they are demographics or latent constructs)
any published work would be of great help. or video illustrating the same.
thank you

Hi,

It's easy, you only have to connect the endogenous variables to control variables that want to be controlled. Examples of the journal can be seen below

http://www.sciencedirect.com/science/ar ... 2614009664

Kind Regards,

thank you

Hengkov wrote:Hi,
It's easy, you only have to connect the endogenous variables to control variables that want to be controlled. Examples of the journal can be seen below
http://www.sciencedirect.com/science/ar ... 2614009664

To be clear, do you mean that we create a latent variable (to act as control variable) and connect (point to) GPD?

Then we run PLS Algorithm or Bootstrapping to test its effect?

Thank you.

Yes, it is very simple.

Yes, exactly.

Instead of introducing control variables, a recommended approach is to consider the respondents' profiles.

Based on the demographics, separate into different groups. For example, gender, age, location, friends, etc.

Thereafter, run PLS-MGA to check for significant differences.

Using a multi group (MGA) approach is more like testing if the controls act as moderators. It evaluates if there is heterogeneity in the effects of your other variables. This is something different, but maybe also very usefull.

Based on using PLS-MGA to determine the significance of relationships across the respondents' profiles, I was able to develop this matrix to give a visual interpretation on determining significant effects based on demographics.

As explained below...

Multigroup Analysis Matrix
In summary, Table 4.55 below shows the Multigroup Analysis Matrix (MAX) is used to determine the significant effects of relationships across the demographics of respondents. For example, in order to achieve the MAX significant effect of anticipated social action on actual social action, it is recommended to target male respondents between the age group of 30-49. Likewise, for MAX significant effect of network externalities on anticipated social action, the demographics of males from Singapore with less than 300 Facebook friends, will yield the most significant result. This finding makes intuitive sense considering the nature of actual social action compared with anticipated social action and network externalities.

Dear Dr. Becker

Will you kindly comment on the discussion found at viewtopic.php?f=5&t=3901 ?

Both of us are attempting to apply PLS-POS into our study.

Thank you.

Hi. Watch my lecture "Dealing with Control Variables in PLS Path Model using SmartPLS: Path Analysis and Interpretation".

viewtopic.php?f=23&t=29351