Correlations & Significance + moderating effect for different scenario's

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
mnieboer1989
PLS Junior User
Posts: 1
Joined: Mon Apr 22, 2019 11:24 am
Real name and title: Marijn Nieboer

Correlations & Significance + moderating effect for different scenario's

Post by mnieboer1989 »

Hello,

I'm no expert in statistics but I will try to explain as good as possible.

My research model consists of 4 different variables:
1. Benefits (Non material vs. material) > Independent variable
2. Perceived risk > Independent variable
3. Type of information (Low vs. High) > moderation variable
4. Willingness to share information > Dependent variable


I collected data by making use of a scenario based survey that consists of 4 scenario's:
1. Material benefits - Highly sensitive information
2. Material benefits - Low sensitive information
3. Non material benefits - Highly sensitive information
4. Non material benefits - Low sensitive information

In my data set I created two new indicators:
  • Benefits: 0 = Non material and 1 = Material
  • Type of information: 0 = Low sensitive information and 1 = High sensitive information

Now I have multiple questions regarding two main topics:
1. I want to know if the different types of benefits have an positive correlation with Willingness to share information to test these hypothesis:
  • H2a: Material benefits have a significant positive effect on the willingness to share information
  • H2b: Non material benefits have a significant positive effect on the willingness to share information
If I run the PLS Algorithm I see a positive correlation between Benefits and Willingness to share information of 0.222 and by running the Bootstrap algorithm it shows that this effect is significant (T = 3.401).
  • Now I struggle to understand if this correlation + T-value counts for the Material Benefits (1) > Willingness to share information relation or the Non material benefits (0) > Willingness to share information relation ? I guess the material benefits because the factor loading to the benefit indicator is 1.0 but i'm not sure if that makes sense.
  • And, if it the answer is Material benefits, how do I find out what the correlation (+ T-value) for the Non material benefits > Willingness to share relation?
'

2. I want to test if the type of information (high vs. low) has a moderation effect:
  • H3a: The type of information has a significant moderating effect on the correlation between material benefits and willingness to share with the result that the correlation will be less positive if the type of information is Highly sensitive.
  • H3b: The type of information has a significant moderating effect on the correlation between non material benefits and willingness to share with the result that the correlation will be less positive if the type of information is Highly sensitive.
I created a moderating effect by clicking the DV (Willingness to share info) and selecting 'Create moderating Effect' and then select 'Moderation variable = Type of information' + 'Predictor variable = Benefits'. I now see a new variable called 'Benefits * Type of information' and the correlation with Willingness to share information shows a value of 0,1 and the T-value is 1,3 so the relation is not significant. Once again I have the same type of questions how to interpret these results:
  • Are these values applicable for the moderation effect of Highly sensitive information (1) or Low sensitive information (0)?
  • And is, if the answer is Highly sensitive information, the moderation effect tested on the relationship between Material benefits (1) > Willigness to share information or Non material benefits (0) > Willingness to share information?
  • Lastly, if the answers to the question above is Highly sensitive information and Material benefits, how do I get the results for e.g. Low sensitive information and material benefits?
I hope there are people who can help me out, would be much appreciated!

Greetings

Marijn Nieboer
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Correlations & Significance + moderating effect for different scenario's

Post by jmbecker »

If you have a dummy coded (0/1) variable the regression effect of this variable is usually the mean difference in the dependent variable between the two groups conditional on all other variables in the regression. For example, you have a positive 0.3 effect that mean that your DV is 0.3 larger for those with a 1 on your dummy variable than those with a 0. So a positive effect means an increase in the DV for the group with a 1.

1)
This general interpretation is also valid in PLS. However, the coefficient is not directly interpretable as a mean difference, because PLS routinely standardizes all variables (including single-item constructs). Thus, you would need to unstandardize the effect to facilitate this direct interpretation.

So for you example you can conclude that material benefits have a higher willingness to share information given your positive effect of 0.222, but you cannot conclude that willingness to share is on average 0.222 higher for material benefits than non-material benefits. You would need to take the group imbalance (standard deviation of the dummy variable) into account and restandardize the effect. This unstandardization is for example part of the IPMA analysis.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Post Reply