Master thesis: Model evaluation

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
lisakhlr
PLS Junior User
Posts: 1
Joined: Sun Jul 17, 2022 12:13 pm
Real name and title: Lisa Köhler (MA)

Master thesis: Model evaluation

Post by lisakhlr »

Hello All,
I am currently writing my master thesis on nudging and I am a bit overwhelmed with the statistical analysis of my results. Structural equation models were not familiar to me so far and I am still trying to find my way in the jungle.

Description of my procedure and variables:
I conducted an experiment within a company with the goal of getting users to fill out various profile information.
The completed profile attribute is my dependant varaible (binary coded;0=not completed,1=profile completed).

Participants completed three questionnaires. These were 5-point Likert scale from "Strongly agree" to "Strongly disagree."
One questionnaire measured susceptibility to nudging, which would also be my independant variable.
One questionnaire dealt with personality (measured with the Big5). The traits are supposed to predict my independant variable. Thus my hypotheses.
One questionnaire measures Internet Privacy Concerns. This is meant to serve as a control variable to check whether individuals who are more likely to be concerned, about their personal data, generally simply do not provide their data.

In addition, the participants (N=5,183) were divided into 5 different groups within the experiment, each of which was presented with a different stimulus (nudge) that was intended to influence the completion of the profile attributes differently. My nudges were designed differently in terms of intensity (i.e. how strongly is the person influenced?) and tailoring (i.e. how well is the nudge tailored to the person?). I crossed these two characteristics in the design, resulting in these possible pairs:
Group1: High Intensity + High Tailoring.
Group2: High Intensity + Low Tailoring
Group3: Low Intensity + High Tailoring
Group4: Low intensity + Low tailoring
Group5: Control group (did not see a nudge).

I have also already designed a model that describes the relationships as hypothesized. See image here: https://imgur.com/oBfWCq8 (Sorry I am not good with the html code)

Is a structural equation model appropriate here and how do I apply that? Do you guys have any tips on how best to code my nudges in my excel sheet so that I can use them in SmartPLS as an indicator for groups?
Right now I have a bunch of different dummy variables that are binary coded. I tried to use those variables in a multigroup analysis within SmartPLS but I am getting no results only error messages.

Also, I have another key problem: I calculated the Cronbach's Alpha values of my scales, coming up with a CA=.38 for the scale for my UV. So not nearly acceptable. Even by combining and omitting different items within the scale I cannot get above a value of .59. Since this is quite problematic, the consequence is actually that I have to throw this variable out of my model (it should be said that the data have been well cleaned and it is definitely a problem with the scale). Can I still calculate a structural equation model or is another method better suited?

I have SmartPL4 available for the analysis.

I hope I haven't overwhelmed you with this huge post and someone can help me!
Thank you very much for your help :)
samolu
PLS Junior User
Posts: 1
Joined: Mon Sep 12, 2022 2:08 pm
Real name and title: Sam Olu

Re: Master thesis: Model evaluation

Post by samolu »

Hi Lisakh,
Great job you have done and I must commend your effort at the master research level. Your SP must be proud of you. While I haven't gotten much grip on the Smartpls to advise you on how best to proceed, I think you can check this youtube https://www.youtube.com/watch?v=1Vym0flo3fo or search for research HUB channel for ideas.
I used Co-variance analysis in the past and could tell how difficult it was, especially when you posted in forums and no one replied. Don't let that dampen your spirit but use that to learn and know everything you need to scale. Hopefully this will help you a bit

Disclaimer: I am not affiliated to the youtuber but I am just here to assist in my little way.
Post Reply