Effect size interpretation

Posted on
No user picture. disciple Joined: 05/17/2021
Forums
Dear Dr. Cheung,

I am using the calEffSizes function from your metaSEM package to calculate effect sizes for a cross-lagged model with two variables (one is continuous from 1 to 5 and the other is categorical from 0 to 4) across three time points. I am not sure if my code is correct and so could you help me check it? Assuming the code is correct, how do we interpret those effect sizes? Are they similar to Cohen'd? Please find the data in the attachment. Thank you!

Best,

Chao

data(pcdata_comb)
obs.vars1 <- c("B1R_EREG", "mild1",
"B2R_EREG", "mild2",
"B3R_EREG", "mild3")
my.cor1 <- cov(pcdata_comb[,..obs.vars1])
my.n <- nrow(pcdata_comb)
pcmod_6_1 <-'
mild2 ~ a*B1R_EREG + b*mild1
mild3 ~ e*B2R_EREG + f*mild2
B2R_EREG ~ c*B1R_EREG + d*mild1
B3R_EREG ~ g*B2R_EREG + h*mild2
B2R_EREG ~~ i*mild2
B1R_EREG ~~ j*mild1
Dir1 := a
Dir2 := b
Dir3 := c
Dir4 := d
Dir5 := e
Dir6 := f
Dir7 := g
Dir8 := h
Dir9 := i
Dir10 := j'
calEffSizes(model=pcmod_6_1, n=my.n, Cov=my.cor1, lavaan.output=FALSE)
#I got the following output
$ES
Dir1 Dir2 Dir3 Dir4 Dir5 Dir6 Dir7 Dir8 Dir9 Dir10
0.002904291 0.214657782 0.322841276 0.031265807 0.076692156 0.281432562 0.339029029 0.008322744 0.034570783 0.014019034

Replied on Tue, 05/25/2021 - 01:23
Picture of user. Mike Cheung Joined: 10/08/2009

Dear Chao,

As shown in the attached figure, your model is a standard path model. Your effect sizes (Dir1 to Dir10) are just path coefficients (a to j). You may refer to SEM textbooks (e.g., Kline, 2016) to see how to interpret them.

Kline, R. B. (2016). Principles and practice of structural equation modeling (4th ed.). Guilford Press.

Best,
Mike