How do I perform my sequential analysis

[Almostasmartstudent]

Hey all,

At the moment I'm busy with my master thesis, but I am a bit stuck in at my analysis plan. I'm measuring the effect of political oriëntation (my main X) on vaccination readiness (y). In my plan I start with my control variables (C) and my Y-variabele. In the second step I have to add all my other X-variables; 6 in total. But after this step I'm lost. I have to test hypotheses which contain moderators and mediations. All of these M-variables are my X-variables. In step 2 I'm looking for the main effect, In the next step (3) I want to check how these 5 x-variables affect the relationship between the main X (Political Oriëntation) and my Y. So I don't know how to perform step 3 (and/or 4). In the last step I check the interaction-effect.

Who can give me some advice?

 

[Karabiner]

I am not quite sure what you are asking here. If you want to "check how these 5 x-variables affect the relationship between the main X (Political Orientation) and my Y", then this would be done by including the 5 interactions between x1 to x5.and X. But in the last step again you mention "the" (just 1?) interaction effect.

Have you ever consulted Andrew Hayes' Moderation & Mediation site? He has programmed the according SPSS syntax and provides explanations.

With kind regards

Karabiner

 

[Almostasmartstudent]

This is my conceptual model and my question is: How can I test this in SPSS? I have not included my controle variables in this model (yet).

I have learned at the university that step 1 is to add the control variables at the same time in Block one.
Step 2 is to add all the X'es
Step 3 is adding the moderationvariable(s)
Step 4 is adding the interactionvariables.

But in my thesis I'm so lost. Because my model contains both mediations and moderations. Also there is also one arrow which goes both ways (between Sociale value orientation and political orientation).

Can you/someone help me with the question: How can i put this in SPSS. I'm using lineair regression.

 

[Karabiner]

What kind of mediation is this? Do you think that is some direct effect of political preference on readiness, plus some indirect effect which is mediated by social values orientation, plus some other indirect effect mediated by trust?

I am not sure how to tests all this (is sample sze large enough for so many parameters?), especially the mix of mediation and moderation.

Maybe you should consider to do separate analyses, one with the moderations, one with the mediations.

Or maybe you leave out the direct effect of political preference and construct a moderated mediation model (i.e. something like "the effect of orientation on readiness is mediated by trust, but the strength of association between readiness and trust is moderated by social integraton [[or some other variable]")?

The Hayes Macros (for SPSS, Stata and R) would permit the analysis of such simpler models, AFAIK.

With kind regards

Karabiner

 

[Ichbinaarnout]

I failed to log in on my other account unfortunately, So I couldn't go back earlier for a response. But I made some alterations to the model (see below). And it is a partial mediation (if that answers your question) with N = 1345. I will give a little more context of the variables:

X1 - SVO is measured from 0 to 5. A value of 5 means very pro-social and a 0 means not pro-social at all.
Hypothesis 3a: When people are more pro-social, they have a bigger vaccination readiness.
Hypothesis 3b: The effect of the social value orientation on vaccination readiness is for a part explained by someones political preferences.

X2: Institutional Trust is measured from 0 to 10. A value of 10 means that the respondent trusts the government and institutions fully and a 0 means that they don't.
Hypothesis 4a: When people score higher on Institutional, they have a bigger vaccination readiness.
Hypothesis 4b: The effect of the Institutional Trust on vaccination readiness is for a part explained by someones political preferences.

X3 - Social integration is measured from 0 to 1. It is not a binary variable since it is an average of 65 binary items. So it is now a continuous scale. a 0 means that the respondent don't participatie in civil society – so they are not socially integrated. A 1 means that they participate a lot.
Hypothesis 5a: When people score higher on social integration, they have a bigger vaccination readiness.
Hypothesis 5b: The effect of the social integration on vaccination readiness is for a part explained by someones political preferences.

M1 - this is my main focal point of the thesis. It is measured on a 10 point scale where 10 means that the respondent is extremely left and 0 means that they are extremely right. Left and right refers to the political landscape – of-course.
Hypothesis H3c: The more left-wing a person's political affiliation, the greater his willingness to vaccinate.

W - This is a moderator variable. Measured from 0 - 3, but again with values in between sine it is an average of 6 items. 0 means that the respondent has a lot to do with barriers (like fear, lack of knowledge, language problems). 3 means that there aren't problems.
Hypothesis 1a: As someone experiences multiple obstacles, the willingness to vaccinate decreases
Hypothesis 1b: The effect of political preferences on vaccination readiness increases as someone experiences less barriers.

Z - Also a moderator variable - Measured from 1 - 5. 1 means that the health status of the respondent is bad and 5 means that the respondent is very healthy. There are no values between the whole numbers – so there are only 5 possible values.
Hypothesis 2a: When someone is more healthy, the vaccination readiness declines.
Hypothesis 2b: The effect of political preferences on vaccination readiness increases as someone experiences less health problems.
_____________________________________________________________________________________________________________________

In the Hayes Macro it looks like you only can use 1 X-variable at the time. So that is a problem since I have 3. But i can to 3 separate ones with model 16 (see also below). But if i use that 3 times, how can i control for the x's that are not part of that model? Do I have to use them as covariates?

And my last addition: X1, X2, and X3 are correlated with each other.

I hope someone can give me some advice on this one.

 

[Karabiner]

Do you need distinct results for the three predictors, or do they jointly represent a single concept/construct? In the latter case,
it might be possible to collapse them into one variable.

Finally, there are always the structural equation models, but unfortunately I am not familiar enough with them to give useful advice.

With kind regards

Karabiner

 

[Ichbinaarnout]

I need distinct results for the three predictors. But do you think I can use model 16 three times with the other x'es as covariates or doesn't that help?

I heard about SEM, but I never used it, so I don't want to use that method.

 

[Ichbinaarnout]

Someone else who knows if I can fix my problem by using the other X'es as covariates? ):