Control Variables + MGA

Research topics can be discussed in this area.
Post Reply
Raul.B
PLS Junior User
Posts: 1
Joined: Tue Jan 12, 2021 10:37 am
Real name and title: Raul Bravo - PhD candidate

Control Variables + MGA

Post by Raul.B »

Dear forum,

This is my first post as I have just discovered this interesting and insightful forum. Also, I started using Smart PLS3 less than a month ago. So, I hope I will find help regarding the questions below. Thanks a lot in advance.

My model consists of 12 constructs, all reflective. Sample size N=178. Following Hair et al. 2017, I used Consistent PLS in my analysis.

I wanted to check the effect of some control variables (CVs) on my final DV: Gender (0=fem, 1=male), Experience (Likert 1-6), Income (8 ranges, Likert 1-8), and Age (continuous).

1) First, I tried to include all the 4 CVs at once (as single item variables, connecting them to DV). Result (PLSc and Bootstrap.c): only Income effect is significant. However, two paths in my inner model changed from significant to not significant. So, I decided to include the 4CVs one by one to see the difference.

2) I included each CV, one by one and ran PLSc and Bootstrap.c. Again, the analysis returned a significant effect of Income, no significance of the other CVs.

Q1-> Is this method right? Should I go with the "one-by-one" method?

Q2-> How should I report this? Should I include the model with all the 4 CVs showing their significance? Or only Income?

Q3-> Which PLSc results should be reported, the ones with or without the CVs? My R squares increased a bit after the inclusion of CV Income, so should I report only the new R2 or include both before and after CV to show the difference? Same thing for some paths and f2, Q2.

3) I also tried MGA in an attempt to see whether my CVs have any impact on the whole model. I got some interesting results. However, I have a problem with the size of my groups:

Age
21-29: 120
30-40: 58

Income
Low: 117
High: 61

Gender
Female: 99
Male: 79

Experience
0-2 years: 64
>2 years: 114

I read the recommendations of (Becker et al. 2013; Hair et al. 2014; Kock and Hadaya 2016, see image) regarding the sample and group sizes. However, I am not sure I understood it well. My model has 4 R squares ranging 0.102 to 0.655. The latter is my outcome's DV and 7 arrows point at it (8 if we count the Income CV).

Q4-> Should I go with the groups above and run my MGA anyway? I also got some interesting results when I include the Country of origin (cultural background), however, same problem: size of groups are 118 (Europe), 30 (Asia), 28 (America).

Very best,
Raul

MGA size.png
MGA size.png (87.15 KiB) Viewed 24546 times
Post Reply