Hi all,
I am currently working on my master's thesis for the study Economics & Business Economics specialized at marketing. The goal of my research is to analyze the effect of several factors on fan loyalty for football clubs to provide marketers insight which need to be enhanced and which ones mitigated. I have a question concerning moderation analysis. Is it better to use a continuous or categorical variable(3 or 4 categories) as a moderator for this study? Relationship length (how long someone perceives himself as a fan of the respective club) is the moderator, while its corresponding hypothesis is 'Relationship length positively influences the attitudinal-behavioral relationship of fan loyalty.
This is my first time posting here, so if you have any questions let me know!
Thanks in advance!
Kind Regards, Bram
The easiest kind of moderator variable is a 2-category comparison, but that is probably not appropriate here. Things become more complicated as you add more categories because that adds more comparisons. For example, with 3 categories you have 3 possible comparisons: AB, BC, and AC. With 4 categories, there are 6 possible comparisons, and so on.
Given your research question, I would recommend working with a continuous moderator, which is a bit more difficult but the techniques are relatively well know. In particular, I suggest doing a Google on "linear by linear" interaction.
But what do you want to test as your moderation? This is your driving force. Best wishes, David Booth
David Eugene Booth I want to test whether the relationship length of a fan with their respective favorite team moderates the attitudinal-behavioral link of fan loyalty. In other words, a fan who supports the club for a longer time should have a higher degree of behavioral loyalty than a relatively new fan, with the level of attitudinal loyalty being equal.
This looks like a psychometric problem. Ask one of them how best to measure your variables but simple is not often the best choice in such an analysis. Once that is finished you choose the moderator on the basis of research questions. Not just simple. Use good research design
Best wishes, David Booth
Bram Ten Barge Moderation is just another name for interaction effects. So, you can add a multiplicative term (relationship length X attitude) to test for this. As I mentioned earlier, interpreting the interaction between two interval level variables can be a bit demanding, but once you learn how, it is straightforward to interpret.
BTW to follow up on Prof Morgan advice do an interaction plot to see what is going on as well as the regression part. There are great examples at Virginia link given in. The attachment below. D. Booth
Check https://statistics.laerd.com/spss-tutorials/dichotomous-moderator-analysis-using-spss-statistics.php
Also check https://www.statisticshowto.com/moderating-variable/
You can used logistic regression where you can use both continuous and categorical values.
First of all, thanks everyone for their input, it helps me a lot! Kiprotich Kiptum Chinaza Godswill Awuchi David Eugene Booth David L Morgan
I will add the multiplicative term of David L Morgan (attitudinal x relationship length) to my regression. Only one small question, I reckon the covariate age cannot be included in my regression due to a high degree of multicollinearity with relationship length, right? The other covariates I want to use are gender and education, which seem okay to me to add in the regression. I'm almost ready to distribute my survey, just want to make sure that I will measure the right covariates :).
Thanks in advance for answering again!
I think you are right be concerned about collinearity between agar and relationship length, but this is an empirical question, so be sure to check the actual correlation.
Yes, I agree with your concern David L Morgan . I'll check the correlation after the data is gathered. At least I got two proper covariates in gender and education. Thanks for all your help again professor Morgan!
Dear Bram,
Better to conceptualize relationship length moderator as a continuous variable since its nature is not discrete and is continuous.
But for analysis I am strongly suggesting you Preacher and Hayes PROCESS macro. It will test the effect of continuous moderation and simultaneously its conditional effects i.e., High relationship length, medium and low. For these conditions you will also get lower and upper level of conf.....
The variable can be continuous or categorical. The important thing is that there is a third variable that changes the nature of the relationship between the predictor and outcome
- Badges
- Science topic
- Similar topics
- Thesis Research