Help needed: Mediation Analysis

[Boris Cornelis]

Hi guys,
I could really use your help, I am very new with statistical analysis, however need to do a mediation analysis for a research paper.
I read a lot about SEM, and installed the package on Stata.

Only Issue is that my model includes

IV (categorical) = 3 layers/3 conditions = No Non-verbal cues, Positive Non-verbal cues, Negative non-verbal cues
Mediation = Credibility measured using 6 items = I do have an computed average for this
Dependent variable = I have 3 = propensity to invest, amount willing to invest, percentage/equity

So with that said, not entirely sure how to build this model correctly in SEM, given the situation with 3 manipulated conditions with the IV
And the mediation situation whereby 6 items measure the mediation = should I then just use the average within the SEM model?

Any help on how to approach the problem = honestly would help me greatly

[Felix Bittmann]

Yes, it it is feasible create one continuous mediating variable. Did you try to build an index using alpha? If the theory supports it and alpha is high, this is fine. Regarding the main question, this tutorial fits you well: https://stats.idre.ucla.edu/stata/fa...l-iv-in-stata/

[Boris Cornelis]

Yes, it it is feasible create one continuous mediating variable. Did you try to build an index using alpha? If the theory supports it and alpha is high, this is fine. Regarding the main question, this tutorial fits you well: https://stats.idre.ucla.edu/stata/fa...l-iv-in-stata/

Hi Felix, Thanks for your reply appreciate it!
Build an index using alpha? I don't entirely understand this = my knowledge on statistics is very new

[Felix Bittmann]

Then you will need to have a look at the relevant literature There are many great sources, for example https://www3.nd.edu/~rwilliam/stats2/l23.pdf
What you can try as a first test (after you have read more):

alpha credibility1 credibility2 credibility3 credibility4 credibility5 credibility6

This will create a single score out of your 6 original items. This also depends on the scaling of these so maybe post some more info for us.

[Boris Cornelis]

[QUOTE=Felix Bittmann;n1612871]Then you will need to have a look at the relevant literature There are many great sources, for example https://www3.nd.edu/~rwilliam/stats2/l23.pdf
What you can try as a first test (after you have read more):

alpha credibility1 credibility2 credibility3 credibility4 credibility5 credibility6

Ah cronbach, yes I have already measured this (exactly like how you mentioned it above)

[Felix Bittmann]

Given that the theory agrees, the value is fine. Then you can type

alpha credibility1 credibility2 credibility3 credibility4 credibility5 credibility6, gen(mediator)
histogram mediator

If the distribution is fine you can use this variable as a single continuous mediator in your models.

[Boris Cornelis]

Given that the theory agrees, the value is fine. Then you can type

alpha credibility1 credibility2 credibility3 credibility4 credibility5 credibility6, gen(mediator)
histogram mediator

If the distribution is fine you can use this variable as a single continuous mediator in your models.

I think the distribution looks alright.