Hello,
Currently I am struggling with a model, where I want to test for mediation.
I have a dependent variable which is (unordered) categorical, some control variables and an ordinal independent variable.
Let's say I have 'sport someone does' as independent variable, 'BMI' as mediator and 'hearattacks' as dependent variable.
Furthermore, I have some control variables: country and gender.
If I would do a normal regession, I would use mlogit, since my dependent variable has unordered multinomial outcomes. What can I do to test if BMI is a mediator?
Online I found the options of using SEM, but I'm not sure what kind of regression is underlying this.
And I found: Baron and Kenny (1986) proposed a four step approach in which several regression analyses are conducted and significance of the coefficients is examined at each step.
And I also found paramed, mediate, ldecomp, medeff, gformula...
Is it true that using SEM, your regression is based on OLS? I'm assuming that these methods do something similar as the Baron and Kenny method, looking at the different relationships. I'm just not sure what type of regression is underlying this. If I use the Baron and Kenny method, where do I include my control variables? Is this even possible with a multinomial logit model?
I have never done mediation before, so I'm sorry if it doesn't make sense somewhere down the line.
Thanks in advance.
Currently I am struggling with a model, where I want to test for mediation.
I have a dependent variable which is (unordered) categorical, some control variables and an ordinal independent variable.
Let's say I have 'sport someone does' as independent variable, 'BMI' as mediator and 'hearattacks' as dependent variable.
Furthermore, I have some control variables: country and gender.
If I would do a normal regession, I would use mlogit, since my dependent variable has unordered multinomial outcomes. What can I do to test if BMI is a mediator?
Online I found the options of using SEM, but I'm not sure what kind of regression is underlying this.
And I found: Baron and Kenny (1986) proposed a four step approach in which several regression analyses are conducted and significance of the coefficients is examined at each step.
And I also found paramed, mediate, ldecomp, medeff, gformula...
Is it true that using SEM, your regression is based on OLS? I'm assuming that these methods do something similar as the Baron and Kenny method, looking at the different relationships. I'm just not sure what type of regression is underlying this. If I use the Baron and Kenny method, where do I include my control variables? Is this even possible with a multinomial logit model?
I have never done mediation before, so I'm sorry if it doesn't make sense somewhere down the line.
Thanks in advance.
Comment