No AIC/BIC for one-factor models?

75 views
Skip to first unread message

G.J. Melendez-Torres

unread,
Oct 2, 2018, 4:52:32 PM10/2/18
to lavaan
Here is a possibly stupid question I haven't been able to figure out.  When running the measurement invariance command on a one-factor model with a WLSMV estimator and five groups (because all indicators are ordinal), the resultant scaled chi-square difference test table doesn't generate an AIC or BIC value for any of the invariance models.  I could obviously use CFI and RMSEA for similar purposes but it would be helpful to have AIC/BIC values.  I'm not sure it matters but my code is below:

happy <- 'factor =~ VAR_1 + VAR_2 + VAR_3 + VAR_4 + VAR_5 + VAR_6 + VAR_7'

measurementInvariance(happy, data=master_data, estimator="WLSMV", strict=TRUE, group="grade",
                      ordered=c("VAR_1", "VAR_2", "VAR_3", "VAR_4", "VAR_5", "VAR_6", "VAR_7"))

Any help or explanation as to what kind of surely basic issue I'm missing here would be gratefully received.

Steve Miller

unread,
Oct 2, 2018, 4:54:12 PM10/2/18
to lav...@googlegroups.com
AIC and BIC are based on the likelihood function.   However, you're not using likelihood based estimation with WLSMV and therefore, these are not produced.

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

G.J. Melendez-Torres

unread,
Oct 2, 2018, 5:07:35 PM10/2/18
to lavaan
...this was my first thought yet the problem remains when I use the MLR (or even the ML) estimator.

Terrence Jorgensen

unread,
Oct 6, 2018, 4:53:02 PM10/6/18
to lavaan
 yet the problem remains when I use the MLR (or even the ML) estimator.

Are you still declaring the same variables as ordered?  lavaan 0.6-3 returns an error when you request ML estimation for categorical outcomes.

HS9 <- HolzingerSwineford1939[,c("x1","x2","x3","x4","x5",
                                 
"x6","x7","x8","x9")]
HSbinary <- as.data.frame( lapply(HS9, cut, 2, labels=FALSE) )
fit1
<- cfa(HS.model, data = HS9, ordered = paste0("x", 1:9), estimator = "mlr")


Terrence D. Jorgensen
Postdoctoral Researcher, Methods and Statistics
Research Institute for Child Development and Education, the University of Amsterdam

G.J. Melendez-Torres

unread,
Oct 7, 2018, 9:33:27 AM10/7/18
to lav...@googlegroups.com
Hi, Terrence--thanks for this.  No, I treat all manifest indicators as continuous when I run with ML, viz.:

happy <- 'factor =~ VAR_1 + VAR_2 + VAR_3 + VAR_4 + VAR_5 + VAR_6 + VAR_7'

measurementInvariance(happy, data=master_data, estimator="ML", strict=TRUE, group="grade")

--
You received this message because you are subscribed to a topic in the Google Groups "lavaan" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lavaan/2ZS0dS2S47E/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lavaan+un...@googlegroups.com.

To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.


--
G.J. Melendez-Torres, RN
UK: +44 (0) 7847 360259
melendez...@gmail.com

Terrence Jorgensen

unread,
Oct 11, 2018, 5:57:18 AM10/11/18
to lavaan

 yet the problem remains when I use the MLR (or even the ML) estimator.

Well, using the latest versions of lavaan (0.6-3) and semTools (0.5-1), both AIC and BIC show up in the "Chi Square Difference Test" part of the output.  If you want to see them in the "Fit measures" part, you have to request them to override the default c("cfi","rmsea")

HW.model <- ' g =~ x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 '
measurementInvariance
(model = HW.model, data = HolzingerSwineford1939,
                     
group = "school", fit.measures = c("aic","bic"))

Reply all
Reply to author
Forward
0 new messages