DWLS as an estimator

[Er. OU]

Hi

Variables of my data set doesn't show multivariate normality. N =220. I have tested the same model with ML and DWLS  However both model works fine. DWLS fit indices are better. I have read some papers which gave me the idea to use DWLS. I read some posts which Yves suggested to use robust technique instead of DWLS. I would like to know whether there is a general tendency for it or it depends on the cases? If I don't encounter any error neither for ML or DWLS. Is there any reason not to use DWLS.

Any answer will be appreciated.

Regards.

[Er. OU]

My second question is about AIC and BIC. As far as I know AIC and BIC is specific for ML but not for other estimators. Are there any estimator specific for DWLS.

I know that it is important to compare AIC and BIC for model and saturated ones but I couldn't find model AIC in the output. There are just one AIC an one BIC value.

Regards.

28 Kasım 2014 Cuma 15:54:33 UTC+2 tarihinde Er. OU yazdı:

[Edward Rigdon]

     In a 1969 paper in Psychometrika, Karl Jöreskog proved analytically that ML is most statistically efficient--when assumptions hold.

--Ed Rigdon

--
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 http://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[Er. OU]

Thanks Edward 

I glanced on the paper you referred to but haven't seen any comparison in between the techniques but I know that ML is the most preferred one. For my case normality assumption doesn't hold, I have read some recent papers about the appropriateness of DWLS. Do you think to use it is reasonable or not? If not, is there any method which should prefer instead of DWLS rather than ML

Regards.

28 Kasım 2014 Cuma 16:46:35 UTC+2 tarihinde Edward Rigdon yazdı:

[Seongho Bae]

Find out this:

Kline, R. B. (2011). Principles and practice of structural equation modeling. Guilford press.

DWLS with scaled.shifted test (a.k.a. WLSMV) may better for categorical data like likert type scale when use ordered option in lavaan. Cheers.

--

Seongho

2014년 11월 29일 토요일 오전 1시 45분 6초 UTC+9, Er. OU 님의 말:

[Er. OU]

Hi Seongho

Which page are you refering to. I have the same book as third edition 2011. I had searched the book and only found the following for DWLS

Another option in LISREL is diagonally weighted least squares

(DWLS) estimation, which is a mathematically simpler form of WLS estimation that 

may be better when the sample size is not very large. 

Regards

28 Kasım 2014 Cuma 21:40:03 UTC+2 tarihinde Seongho Bae yazdı:

[Seongho Bae]

See 176~183 pages.

WLSMV is DWLS too.

2014년 11월 29일 토요일 오전 5시 3분 45초 UTC+9, Er. OU 님의 말:

[yrosseel]

I do not think there is a lot of literature out there comparing DWLS vs
robust-ML when data is continuous but non-normal, and the sample-size is
small...

Yves.

[Er. OU]

Thank you Yves

Is it applicable to report the results using DWLS. I would like to know whether my analysis will be applicable or questionable because of the DWLS technique based on your experience and expertise

Regards.

1 Aralık 2014 Pazartesi 10:01:36 UTC+2 tarihinde yrosseel yazdı:

[yrosseel]

On 12/01/2014 09:53 AM, Er. OU wrote:
> Thank you Yves
> Is it applicable to report the results using DWLS. I would like to know
> whether my analysis will be applicable or questionable because of the
> DWLS technique based on your experience and expertise

The standard nowadays is robust ML. Without further published evidence
in favour of DWLS (for continuous non-normal data), I would not use it
(yet) for publication purposes.

Yves.

[Er. OU]

Thanks a lot for your help.

1 Aralık 2014 Pazartesi 15:42:36 UTC+2 tarihinde yrosseel yazdı:

[mu peker]

copy-paste from https://www.ncbi.nlm.nih.gov/pubmed/27571021

Psychol Methods. 2016 Sep;21(3):369-87. doi: 10.1037/met0000093.

The performance of ML, DWLS, and ULS estimation with robust corrections in structural equation models with ordinal variables.

Li CH ¹ .

1

National Sun Yat-sen University.

Abstract

Three estimation methods with robust corrections-maximum likelihood (ML) using the sample covariance matrix, unweighted least squares (ULS) using a polychoric correlation matrix, and diagonally weighted least squares (DWLS) using a polychoric correlation matrix-have been proposed in the literature, and are considered to be superior to normal theory-based maximum likelihood when observed variables in latent variable models are ordinal. A Monte Carlo simulation study was carried out to compare the performance of ML, DWLS, and ULS in estimating model parameters, and their robust corrections to standard errors, and chi-square statistics in a structural equation model with ordinal observed variables. Eighty-four conditions, characterized by different ordinal observed distribution shapes, numbers of response categories, and sample sizes were investigated. Results reveal that (a) DWLS and ULS yield more accurate factor loading estimates than ML across all conditions; (b) DWLS and ULS produce more accurate interfactor correlation estimates than ML in almost every condition; (c) structural coefficient estimates from DWLS and ULS outperform ML estimates in nearly all asymmetric data conditions; (d) robust standard errors of parameter estimates obtained with robust ML are more accurate than those produced by DWLS and ULS across most conditions; and (e) regarding robust chi-square statistics, robust ML is inferior to DWLS and ULS in controlling for Type I error in almost every condition, unless a large sample is used (N = 1,000). Finally, implications of the findings are discussed, as are the limitations of this study as well as potential directions for future research. (PsycINFO Database Record.

1 Aralık 2014 Pazartesi 20:47:45 UTC+2 tarihinde Er. OU yazdı:

[joh4nd]

Do you agree that as of today, almost 10 years later, DWLS is recommended for continuous (from polychoric correlations of ordinal data) non-normal data?