new release: estimator = MLM gone?

[Pahud]

Dear lavaan

If I specify estimator = "MLM" I get the following errmsg:

Error in t(Delta[[g]]) %*% A1[[g]] : non-conformable arguments

This started with the lavaan update... before everything was fine. Running the identical code without any changes.

Thanks,

Olivier

[Pahud]

I saw the workaround proposed by yves:

 fit.model <- sem(model,
                        data      = data,     
                        estimator = "ML",
                        se = "robust.huber.white",
                        test = "yuan.bentler",
                        mimic     = "Mplus")

However, this won't lead to identical results as before and no CFI is reported (i.e., NA shown).

[Amonet]

I had some issue with the MLR estimator also, and as you said: different results from the previous version (0.5-23-1097). 

For this reason, I am currently still using the previous version. If you're interested, you can do so also using the following: 

library(devtools)

install_version("lavaan", version = "0.5-23.1097", repos = "http://cran.us.r-project.org")

kind regards,

amonet

[Yves Rosseel]

On 07/13/2018 02:30 PM, Pahud wrote:
> Dear lavaan
>
> If I specify estimator = "MLM" I get the following errmsg:
> Error in t(Delta[[g]]) %*% A1[[g]] : non-conformable arguments

Could you give us a reproducible example? I have not been able to
reproduce this (yet).

The 'workaround' you refer to was only for the robust CFI/RMSEA values,
and only if estimator = "MLR".

Yves.

[Pahud]

Hey Yves

I "downdated" my version and it works with any flaws.

With the newest version (0.6-1) it doesn't matter what model I specific or what data I use. I tried several models (that I have been working with for over a year now and never experienced that type of problem) and it always comes down to the MLM estimator and not to the code/data. If I change the estimator (e.g. to ML) everything works fine.

The 'workaround' you refer to was only for the robust CFI/RMSEA values,

and only if estimator = "MLR".-> ok, then this isn't the workaround for MLM -> hence, different results.

[Yves Rosseel]

On 07/16/2018 09:03 AM, Pahud wrote:
> With the newest version (0.6-1) it doesn't matter what model I specific
> or what data I use. I tried several models (that I have been working
> with for over a year now and never experienced that type of problem) and
> it always comes down to the MLM estimator and not to the code/data. If I
> change the estimator (e.g. to ML) everything works fine.

Please, can you give us a reproducible example? Or at least show us the
function call? I can not reproduce this, but I suspect this is due to
some combination of arguments... I can only fix it if I understand what
is going on.

Yves.

[ja sp]

I got the same errormsg... Error in t(Delta[[g]]) %*% A1[[g]] : non-conformable arguments

I used the following funciton:

BRD <- 'BRDg =~ S1mean +S2mean + S3mean + S4mean'

fit.BRD <- sem(BRD,
             data = data,
             estimator = "MLM",
             mimic = "Mplus")

fitmeasures(fit.BRD, c("npar", "chisq", "df", "cfi", "rmsea", "srmr"))

Find a small exemplary sample attached.

Thanks,

Jasmin

[Yves Rosseel]

Fix: either add meanstructure = TRUE, or remove mimic = "Mplus"

Yves.

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

[Yves Rosseel]

Fixed in 0.6-4.1334

Yves.