weighted, unweighted and unit weights for composite reliability

42 views
Skip to first unread message

Balal Ezanloo

unread,
May 21, 2020, 6:48:34 AM5/21/20
to lavaan
Greeting every one 

I need to compute composite reliability based on Jöreskog (1971). i know that maximalRelia function in semTools compute some kind of composite reliability. but i do not  know:
1) how it determine weights for items? 
 
2) how we can compute composite reliability based on weighted items, unweighted items and unit weights for all items.

Best

Yves Rosseel

unread,
May 26, 2020, 3:24:41 PM5/26/20
to lav...@googlegroups.com
On 5/21/20 12:48 PM, Balal Ezanloo wrote:
> that maximalRelia function in semTools compute some kind of composite
> reliability. but i do not  know:
> 1) how it determine weights for items?

See the man page

?maximalRelia

and the references therein.

Message has been deleted
Message has been deleted
Message has been deleted
Message has been deleted

Balal Ezanloo

unread,
May 27, 2020, 3:01:24 AM5/27/20
to lavaan
Thanks, i find solution to my question. hear is an example from lavaan to show how maximalRelia computes item weights.


require(lavaan)
require(semTools)
total <- 'f =~ x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 '
fit <- cfa(total, data = HolzingerSwineford1939,std.lv = TRUE)
summary(fit)
lambda <- inspect(fit,"est")$lambda
lambda
theta <- round(diag(inspect(fit,what = "est")$theta),3)
theta
myweigths <- as.vector(lambda/theta) 
myweigths
maximalRelia(fit)
# Extract the weight
mr <- maximalRelia(fit)
maximalRelia.weigths <- attr(mr, "weight")
maximalRelia.weigths

# item weights computed manully. compare them with item weights from maximalRelia()
 round(cbind(maximalRelia.weigths,myweigths),3)
   maximalRelia.weigths myweigths
x1                0.466     0.465
x2                0.197     0.197
x3                0.208     0.208
x4                2.597     2.593
x5                2.236     2.230
x6                2.579     2.575
x7                0.172     0.171
x8                0.208     0.207
x9                0.337     0.336

# compare weighted reliability computed manully with  weighted reliability from maximalRelia()
round(cbind(sum(lambda^2/theta)/(1+sum(lambda^2/theta)),maximalRelia(fit)[1]),3)
      [,1]  [,2]
[1,] 0.887 0.887

 
Reply all
Reply to author
Forward
0 new messages