Hi Mike,
I am trying to conduct a multivariate meta-analysis by using metaSEM. I reviewed your article Cheung, M. W. L. (2014). metaSEM: An R package for meta-analysis using structural equation modeling. Frontiers in Psychology, 5. In the supplementary material, you provided some examples. In terms of the second example in section 3, you used the data set reported by Aloe et al. (2014).
Now, I am a little confused how to examine the mixed-effects model. You replicate the analysis and take Publication_type for example. But I don’t know how to deal with a continuous variable as a moderator such as Years of experience. So could you provide the code for testing Years of experience as a moderator following your script.
Here is the script,
head(Aloe14)
meta1 <- meta(y=cbind(EE,DP,PA),
v=cbind(V_EE, C_EE_PA, C_EE_PA, V_DP, C_DP_PA, V_PA),
data=Aloe14)
summary(meta1)
( coef1 <- coef(meta1, select="random") )
my.cov <- vec2symMat(coef1, byrow=TRUE)
dimnames(my.cov) <- list( c("EE", "DP", "PA"),
c("EE", "DP", "PA") )
my.cov
( cov2cor(my.cov) )
plot(meta1, main="", axis.labels=c("EE", "DP", "PA"))
( journal <- ifelse(Aloe14$Publication_type=="Journal", 1, 0) )
meta2 <- meta(y=cbind(EE,DP,PA),
v=cbind(V_EE, C_EE_PA, C_EE_PA, V_DP, C_DP_PA, V_PA),
x=journal, data=Aloe14)
summary(meta2)
anova(meta2, meta1)
Bets wishes,
Yu Xie
Dear Yu Xie,
It is the same as that for univariate meta-analysis. Here is an example.
Hope it helps.
Regards,
Mike
Hi Mike
Thank you for your response. It helps a lot.
Best,
Yu Xie