Multigroup SEM with std.lv=TRUE

54 views

Skip to first unread message

Anran

unread,

Jul 10, 2019, 8:05:20 PM7/10/19

to lavaan

Hi everyone,

I am running multiple SEM in lavaan. When I set std.lv=TRUE, the model will converge. However, when I leave it as default, the model does not converge. I am relatively new in SEM, could someone please tell me why this should happen?

The codes are as follows,

ALL.model5<-'

External =~ sf1+sf2+sf3+sf4+sq1+sq2

Improvement =~ si1+si2+si3+si4+si5+ti1+ti2+ti3+ti4+ti5+ti6

Irrelevance =~ ig1+ig2+ig3+bd1+bd2+bd3+bd4+bd5

affect=~pe1+pe2+ce1+ce2+ce3+ce4+ce5+ce6

General =~ sf1+sf2+sf3+sf4+sq1+sq2+si1+si2+si3+si4+si5+ti1+ti2+ti3+ti4+ti5+ti6+ig1+ig2+ig3+bd1+bd2+bd3+bd4+bd5+pe1+pe2+ce1+ce2+ce3+ce4+ce5+ce6

General ~~0*External

General ~~0*Improvement

General ~~0*Irrelevance

General ~~0*affect

External ~~0*Improvement

External ~~0*Irrelevance

External ~~0*affect

Improvement ~~0*Irrelevance

Improvement ~~0*affect

Irrelevance ~~0*affect

TM_EFFORT =~ tm_ef1+ tm_ef4 + tm_ef5

TM_Importance =~ tm_ip3 + tm_ip4 + tm_ip5

TM_TA =~ tm_ta1 + tm_ta2

TM_EFFORT ~ TM_Importance

TM_EFFORT ~ TM_TA

TM_EFFORT ~ General

TM_Importance ~ External

TM_Importance ~ General

TM_TA ~ General

TM_TA ~~ TM_Importance

fit.All.model5.config <- sem(ALL.model5, estimator="MLM", data=OriData,

group="TM_Conditions")

and lavaan gives me warning:

Warning message:

In lavaan::lavaan(model = ALL.model5, data = OriData, group = "TM_Conditions", :

lavaan WARNING: the optimizer warns that a solution has NOT been found!

However, if I do

fit.All.model5.config <- sem(ALL.model5, estimator="MLM", data=OriData, std.lv=TRUE,

+ group="TM_Conditions")

lavaan would process normally.

Am I not supposed to use std.lv=TRUE in sem?

Thank you in advance!

Anran

Terrence Jorgensen

unread,

Jul 13, 2019, 12:23:46 PM7/13/19

to lavaan

When I set std.lv=TRUE, the model will converge. However, when I leave it as default, the model does not converge. I am relatively new in SEM, could someone please tell me why this should happen?

When it converges with std.lv=TRUE, do you have any negative factor loadings that would be unexpected? Or any negative residual variances? It is a bifactor model, so I wonder whether the first factor loading for the same item of 2 different factors being fixed to 1 would cause any problems in case the model doesn't actually fit the items that well. Providing your data might help troubleshoot.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam