0

Theoretically I have two competing mediators (M1 and M2), each of which possibly mediates the relationship between my independent (X) and dependent (Y) variable. I implemented two different mediation analyses using the lavaan package in R with bootstrapping technique for the same dataset: one with the mediator M1, and one with the mediator M2. The coefficient estimate for the mediating effect (i.e. the indirect effect a1*b1) of M2 (β=0.199) is higher than that of M1 (β=0.170). How can I find out if the mediating effects of these two mediators are statistically significant? If it is so, it can be concluded that M2 has a stronger mediating effect than M1 on the relationship between X and Y, can't it?

I have tried to find some information on comparing the coefficients of different mediators for the same statistical model, but what I could get was related only to comparing the coefficients of different independent variables for the same or different models but not mediators.

My model is a moderated mediation model, but at the moment what I am interested in is to compare the mediation effects, not the effects of moderated mediation.

Hopefully I am not taking anybody's time in vain.

I would very much appreciate if someone could help me in this issue. Thanks a lot in advance.

Here's my code:

# For Mediator 1 M1

ModMediation1 <- ' M1 ~ a1*RD + a2*PD + a3*RDXPD + a4*FA +
              a5*FS + a6*FG + a7*T
              
              Perf ~ b1*M1 + b2*FA + b3*FS + b4*FG + b5*T +
              c2*PD + c3*RDXPD
              
              # indirect effect
              IndEff := a1*b1
              
              # index of moderated mediation 
              IndModMed := a3*b1
              
            '

ModMediation1_fit <- lavaan::sem(ModMediation1, data = p2_df, se = "bootstrap", bootstrap = 10000)

summary(ModMediation1_fit, fit.measures = TRUE, rsq = TRUE, standardized = TRUE, ci = TRUE)

parameterestimates(ModMediation1_fit, boot.ci.type = "bca.simple", standardized = TRUE, level = 0.95)

# For Mediator 2 M2

ModMediation2 <- ' M2 ~ a1*RD + a2*PD + a3*RDXPD + a4*FA +
              a5*FS + a6*FG + a7*T
              
              Perf ~ b1*M2 + b2*FA + b3*FS + b4*FG + b5*T +
              c2*PD + c3*RDXPD
              
              # indirect effect
              IndEff := a1*b1
              
              # index of moderated mediation 
              IndModMed := a3*b1
              
            '


ModMediation2_fit <- lavaan::sem(ModMediation2, data = p2_df, se = "bootstrap", bootstrap = 10000)

summary(ModMediation2_fit, fit.measures = TRUE, rsq = TRUE, standardized = TRUE, ci = TRUE)

parameterestimates(ModMediation2_fit, boot.ci.type = "bca.simple", standardized = TRUE, level = 0.95)
Mel
  • 1
  • 2
  • Welcome to StackOverflow. It sounds like this question would be better suited for [CrossValidated](https://stats.stackexchange.com/), not StackOverflow. If you have a specific R programming question, please make your post [reproducible](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) by providing your data and code. – jrcalabrese Dec 03 '22 at 18:15
  • Thank you. I added my code into the text. I am sorry for posting my question to an inappropriate platform. Should I post it to CrossValidated then? – Mel Dec 04 '22 at 08:29
  • If you don't have a question about R programming or `lavaan`, I think CrossValidated would a better place. Make sure to also include the output or summary table that you are having trouble interpreting in future questions. – jrcalabrese Dec 04 '22 at 16:59
  • OK. Sorry for engaging the wrong platform. Thank you for the advice. – Mel Dec 05 '22 at 11:20

0 Answers0