Hi everyone, I have applied SEM to a simple model of IV→DV (Six IV's and One DV) to check the affect of various IV on Buying Intention(DV). I know SEM is used for complex behavioral models. can it be applied to simple models also? IF yes than what will be the advantages of that?
Yes, SEM with latent variables can be applied to simpler regression models, and one advantage is that you can take measurement error into account. Ideally, the SEM results would be less biased (because they account for measurement error) than manifest variable regression results.
See also: https://youtu.be/IfVnWD6QxdE
Yes, SEM can be applied to simple models like the one you described. SEM is a flexible statistical technique that allows for the analysis of complex models with multiple variables and relationships, but it can also be used to examine simple models with just a few variables.
Some of the advantages of using SEM for simple models include:
Overall, SEM can be a powerful tool for analyzing simple models, providing researchers with a range of advantages that can help them better understand the relationships between variables in their model.
Hi,
beyond what the others said, please be aware that a regression-type models shares the fate of all saturated models (such as partial mediation models) that is, that they don't contain testable implications. Hence, you could flip around arrows and wouldn't note that. If you specify latent variables with multiple indicators, you will have positive degrees of freedom but this testing portion comes solely from the restrictions (local independences) of the measurement model. The structural model has no df. Hence, you have to simply belief that the causal directions are properly specified.
Good luck
Holger
Hey Huzaifa
Great answers so far - in addition I suppose you get all the benefits of using maximum likelihood estimation, such as robust estimation, and you can handle any missing data in an efficient and easy way. You can also model 2 DVs at the same time. This is compared to doing the regression in SPSS (or something similar).
But if you decide to run the regression model in SEM, maybe using software like Mplus, the real benefit is that you will see how easy it is to extend or change your model to something like path analysis. In SPSS the model is fixed, all you can do is specify the IVs and DV, but you'll soon see that specifying more theoretically pleasing models is quite simple. There are probably thousands of options in the pull-down 'Analyze' tab in SPSS - it's a lifetime's work trying to understand all of the different tests and models. In Mplus you'll be able to run any model you want using only a maximum of 3 words; On, With, By. Take the leap and you'll never look back.
Mark
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression