1. Introduction to latent class analysis
Вставка
- Опубліковано 28 січ 2021
- This video is aimed to serve as the very first introduction to latent class analysis. I explain the different steps and aims of a latent class analysis, and illustrate this with a real data example using the Latent GOLD 6.0 software. The next videos in this series deal in more detail with topics such as the model equations, model selection, classification, and relating latent classes to external variables.
This series is very helpful and comprehensive. Thank you!
thank you
Thank you Dr. Vermunt! It is very useful video tutorial. On my research, I'm trying to use LCCA (Latent class cluster analysis) and considering to buy Latent Gold as appropriate software to do this. I was wondering about in the end of your video when you put gender as covariate, is that type of LCCA? I read some literature that LCCA is also combining Indicator and Covariate.
Thanks!
Yes, it is possible to include covariates in a LCCA. An alternative is to use a 3-step approach, which separates the identification of the classes based on the indicators and the prediction of the classes using a set of covariates. This is explained in video number 9.
Thank you for this, and thank you for Latent Gold.
My question is about the number of indicators vs. number of classes. I read somewhere that the number of classes should not exceed the number of indicators. Is this the case? Many thanks.
There is no clear rules about this. But in cluster-like applications it usually doesn't make sense to have more latent classes than indicators.
Hello Dr. Vermunt - I have a question. In this example, how would one know that a 5-cluster model would not have resulted in a better model fit (entropy R-squared, BIC, etc)?
When BIC turns out to be lowest for the 4-cluster model, one should indeed check whether the 5-cluster model yields an even lower BIC. I specified 1-4 clusters because this seemed to be reasoble starting point given the number of indicators. But it can of course be that the specified maximum of 4 is too small.
I would not use entropy R-squared for model selection, but only as a measure for how well the selected model classifies individuals.
I absolutely love the video. I purchased the software (lifetime purchase). Unfortunately, it does not install at all. Keeps giving error "Automation Object Not Found" error. Contact technical support has not been very effective as of now. I am hoping that my emails will find someone with the capacity to help me.
Is this actually LCA? I was told that SPSS does not have the functionality to run LCA
Indeed SPSS does not have LCA. But this is the Latent GOLD program which has a graphical user interface similar to SPSS, and allows using SPSS data sets.
Where can I get Latent GOLD 6.0?
If you Google "latent gold statistical innovations", it should be the first link on the page.
No need to read the text on the PP presentation. Assume people can read!
no, keep on reading it. I am playing a mindless game on my phone to relax while listening... at another time I might have been doing chores! Some of us don't have the time to be watching as well as listening to a video. I only need to understand the methodology enough to improve my critique of a study.