Hello
I'm learning moderated mediation and I'm hoping someone will help me understand the conceptual aspect of the conditional indirect effects and how they are estimated using the nlcommand.
The example I reference here is model 2 ( continuous moderator affects path a of the mediation model), which has been thoroughly outlined (analysis wise) in the UCLA link below
https://stats.oarc.ucla.edu/stata/fa...tion-in-stata/
For reference:
x= independent variable
y = dependent variable
w = moderator variable
xw = interaction of IV and moderator
m = mediator variable
For this model 2-- the conditional indirect effect is derived as follows:
conditional indirect effect = b1(a1 + a3w)
conditional indirect effects are obtained for moderator variable at the mean, as well as at 1SD above and below the mean.
coef legends for the above variables:
x = _b[m:x]
w = _b[m:w]
xw = _b[m:wx]
m = _b[y:m]
Taking a look at the nlcom code for 1SD below the mean on the moderator
If my understanding is correct-- then path a is everything to the left of _b[y:m].,
which can be conceptualized visually as:
This is where I have questions. Why is the interaction term multiplied and not added in the path a portion of the nlcom command...?
Put differently: Why- for path a, in the nlcom command, am I adding x + w, but multiplying it by XW ?
1. Is it because the x and w are additive effects on M, but the xw is multiplicative? (could be as simple as that...).
I have a feeling it has to do with the original formula for conditional indirect effect = b1(a1 + a3w)
but I have a hard time conceptualizing how that equation above translates to the nlcom command in question.
I plan to explore other models so understanding how these nonlinear combinations are created is paramount.
Any insights would be greatly appreciated.
Best,
Katherine.
I'm learning moderated mediation and I'm hoping someone will help me understand the conceptual aspect of the conditional indirect effects and how they are estimated using the nlcommand.
The example I reference here is model 2 ( continuous moderator affects path a of the mediation model), which has been thoroughly outlined (analysis wise) in the UCLA link below
https://stats.oarc.ucla.edu/stata/fa...tion-in-stata/
For reference:
x= independent variable
y = dependent variable
w = moderator variable
xw = interaction of IV and moderator
m = mediator variable
For this model 2-- the conditional indirect effect is derived as follows:
M = a0 + a1x + a2w + a3xw
Y = b0 + b1m + b2x + b3w + b4xw
conditional indirect effects are obtained for moderator variable at the mean, as well as at 1SD above and below the mean.
Code:
quietly summarize w global m =r(mean) global s = r(sd) generate xw =w*x /* moderator 1 by IV interaction */ sem (m <- x w wx)(y <- m x w wx)
coef legends for the above variables:
x = _b[m:x]
w = _b[m:w]
xw = _b[m:wx]
m = _b[y:m]
Taking a look at the nlcom code for 1SD below the mean on the moderator
Code:
nlcom (_b[m:x]+($m-$s)*_b[m:wx])*_b[y:m] /* mean - 1 sd */
If my understanding is correct-- then path a is everything to the left of _b[y:m].,
which can be conceptualized visually as:
This is where I have questions. Why is the interaction term multiplied and not added in the path a portion of the nlcom command...?
Put differently: Why- for path a, in the nlcom command, am I adding x + w, but multiplying it by XW ?
1. Is it because the x and w are additive effects on M, but the xw is multiplicative? (could be as simple as that...).
I have a feeling it has to do with the original formula for conditional indirect effect = b1(a1 + a3w)
but I have a hard time conceptualizing how that equation above translates to the nlcom command in question.
I plan to explore other models so understanding how these nonlinear combinations are created is paramount.
Any insights would be greatly appreciated.
Best,
Katherine.
