Multiple Mediation with covariates

[Chia-Hao Shih]

Dear all,

I am trying to run a multiple mediation model in lavaan with 1 predictor (X), 4 parallel mediators (M1, M2, M3, M4), and 3 variables that I want to control for (i.e., age, duration, and severity). I wonder if someone could check my syntax and let me know if my model specification is correct. Thanks in advance.

Model <- ' 

      X ~ b1 * M1 + b2 * M2 + b3 * M3 + b4 * M4 + c1 * X + c2 * age + c3 * duration + c4 * severity

     C1 ~ a1 * X + a5 * age + a6 * duration + a7 * severity

     C2 ~ a2 * X + a8 * age + a9 * duration + a10 * severity

     C3 ~ a3 * X + a11 * age + a12 * duration + a 13 * severity

     C4 ~ a4 * X + a14 * age + a15 * duration + a 16 * severity

     indirect1 := a1 * b1

     indirect2 := a2 * b2

     indirect3 := a3 * b3

     indirect4 := a4 * b4

     total := c1 + c2 + c3 + c4 + (a1 * b1) + (a2 * b2) + (a3 * b3) + (a4 * b4)

     C1 ~~ C2

     C1 ~~ C3

     C1 ~~ C4

     C2 ~~ C3

     C2 ~~ C4

     C3 ~~ C4

'

require(lavaan)

fit <- sem(model = Model, data = Data, se = "bootstrap", bootstrap = 5000)

summary(fit, estimates = TRUE, ci = TRUE)

[Christopher David Desjardins]

Hi, 

This isn't quite right. If M1 though M4 are mediators then need to be on the left-hand side. Also, is each mediator associated with a single C outcome? I'm assuming that below:

Model <- '

# Mediator is regressed on to X

M1 ~ b1*X + age + duration + severity

M2 ~ b2*X + age + duration + severity

M3 ~ b3*X + age + duration + severity

M4 ~ b4*X + age + duration + severity

# Y (Cs) is regressed onto mediator and X

C1 ~ c1*X + b1*M1 + age + duration + severity

C2 ~ c2*X + b1*M2 + age + duration + severity

C3 ~ c3*X + b1*M3 + age + duration + severity

C4 ~ c4*X + b1*M4 + age + duration + severity

indirect1 := a1 * b1

indirect2 := a2 * b2

indirect3 := a3 * b3

indirect4 := a4 * b4

totalC1 := c1 + (a1 * b1)

totalC2 := c2 + (a2 * b2)

totalC3 := c3 + (a3 * b3)

totalC4 := c4 + (a4 * b4)

'

But I am just guessing this is what you want ...

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[Chia-Hao Shih]

Hi Chris,

Thanks for your reply. I realized that I made typos on the original syntax: the C1 to C4 (capital C) should be M1 to M4 and X should be Y. Also, based on your reply, I guess I don't need to assign coefficients to my covariates. Below is my revised syntax, hope you or other people in the group could give me feedback. Thank you.

Model <- ' 

      Y ~ b1 * M1 + b2 * M2 + b3 * M3 + b4 * M4 + c1 * X + age + duration + severity

     M1 ~ a1 * X +  age +  duration +  severity

     M2 ~ a2 * X +  age +  duration +  severity

     M3 ~ a3 * X +  age +  duration +  severity

     M4 ~ a4 * X +  age +  duration +  severity

     indirect1 := a1 * b1

     indirect2 := a2 * b2

     indirect3 := a3 * b3

     indirect4 := a4 * b4

     total := c1 + (a1 * b1) + (a2 * b2) + (a3 * b3) + (a4 * b4)

     M1 ~~ M2

     M1 ~~ M3

     M1 ~~ M4

     M2 ~~ M3

     M2 ~~ M4

     M3 ~~ M4

[Christopher Desjardins]

This looks fine to me. It’s not necessarily clear to me that you need to allow your mediators to covary. So, you could remove that. Do you have a reason or a citation as to why you are covarying them?

You might also be interested in the size of each indirect effect. For example, maybe you want this in your code?

a1b1 := a1*b1

a2b2 := a2*b2

If you think the effects are the same for each regression, then you can assign labels to age, duration, and severity. Otherwise, it’s fine to leave them without the labels.

Chris 

[Chia-Hao Shih]

Hi Chris,

I really appreciate your help. As you pointed out, I don't need my mediators to covary with each other. I saw a sample lavaan syntax online and didn't think it through. Thank you again.

Best,

James