I have road Accidents data (for every variable there is 0 or 1 (eg. If the weather is foggy then it is 1, if it is not foggy than 0).
With ECA, CFA and SEM I do not have too much problems - unless it is a case of dichotomous variables. Can anyone help? Since i'm using AMOS (and not LISREL or MPLUS).....
Thank you
Sincerely,
Darja
Hello
In this site you will find some ppts (#9 and 10) about how to run a binary-outcome SEM in AMOS, using Markov Chain Monte Carlo simulation (MCMC).
http://www.structuralequations.com/AmosTutorials.html
However, i dont find (yet) fit measures with this option.
It's possible with Bayesian Estimation, but I would use Latent Class Analysis (LCA) to identify unobservable subgroups (profile) within a population. I think it's closer to your goal.
The profile (Latente Class) may indicate the characteristics of road accidents. For example, a model with two profiles (two Latent Class) may represent a road accident profile (fog, old tires, snow, high speed, cell phone, gender x, age x, etc.), and a security profile (not foggy, new tires, good weather, average speed, no cell phone, gender y, age y, etc. ). The R package poLCA works with LCA, and is easy to use. This paper shows poLCA in use.
http://www.jstatsoft.org/v42/i10/paper
The following articles may help
Ahmad M. Muayyad (2010). Validation of the Cognitive Appraisal Health Scale with Jordanian patients. Nursing & Health Sciences, 12 (1), 74-79.
Dardas L., & Ahmad M. (2014). Validation of the World Health Organization’s Quality of Life Questionnaire with Parents of Children with Autistic Disorder. Journal of Autism and Developmental Disorders, 44 (9), 2257-2263. DOI 10.1007/s10803-014-2110-1
Define the graph model but use Asymptotically distribution-free estimator (ADF) (View - Analysis Properties - Estimation - Asymptotically Distribution Free).
I would also recommend free R+ program which has nice "lavaan package" with some usefull estimators like DWLS (for binary and ordinal variables). It also has some useful robust estimators like Satorra-Bentler or Yuan-Bentler estimators.
More information is here:
Rosseel, Y. (2013). The lavaan tutorial. Ghent, Belgium: Department of Data Analysis Ghent University.
With regards,
Thank you all.
At this time it is still only “rough idea”. If you are interested in participating with me – with this problem – you are welcome J
Darja
The youtube channel of professor Gaskin was pretty useful for me doing SEM:
https://www.youtube.com/user/Gaskination
Hello
In this site you will find some ppts (#9 and 10) about how to run a binary-outcome SEM in AMOS, using Markov Chain Monte Carlo simulation (MCMC).
http://www.structuralequations.com/AmosTutorials.html
However, i dont find (yet) fit measures with this option.
Macias
I would like to thank You. The link is reallyVERY useful
with regards,
Better If you use Bayesian Belief Network (BBN) model instead of SEM...
After an initial run u can reflect your domain knowledge into BBN for a good fit same as in SEM (AMOS).
Maybe using another sofware may be useful. Have you tried?
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling