The whole data was collected on Likert scale. I am using Prof. Gaston's plspm package in R for SEM modeling. I think different age groups must have differences. But, I can't test differences of more than than two groups at structural level. My question is, if I divide the data into various age based subgroups and prepare SEM models separately for each subgroup. Is it meaningful? How to justify the models are significantly different? I was not able to perform ANOVA test to check the difference among models. What should I use? Please guide me Thanks
Dear Farrukh Baig
This is an interesting question. Theoretically, you may build up the the three models based on the age groups, however, that is not going to tell much.
My suggestion for you is the following, you should have a base group (reference), which is so often coded with 0 value and two other groups and you perform one comparison at once.
Here is an example to elaborate on this suggestion. Let's assume that the groups are the following:
1. Young aged respondents
2. Middle age respondents
3. Old age respondents
And the theory of question is their intention to use a new public transportation service.
Then, you would build up your own hypothesis which can that old age are more likely to use the new service and you may confirm this by checking the beta for that path. Assuming that it went as expected. Then you may compare middle to old and later the young with old group.
Personally, I think it would be easier to interpret while performing one comparison at once.
Best Regards,
Belal Edries
1. If you group age then you are throwing away useful information.
2. If you ALWAYS use someone else's programs then you are bound by what that person wanted to know. R has multiple SEM packages and can after all be done as a regression(see @Holger Steinmetz reply here: https://www.researchgate.net/post/What-is-the-difference-between-a-regression-analysis-and-SEM) thus you can study YOUR research( https://www.google.com/search?client=firefox-b-1-d&ei=jF_IX5PXLuWbwbkPtrWG8AQ&q=Structual+equation+modeling+in+R&oq=Structual+equation+modeling+in+R&gs_lcp=CgZwc3ktYWIQAzIHCAAQyQMQDTIECAAQDTIECAAQDTIECAAQDTIECAAQDTIECAAQDTIECAAQDTIECAAQDTIGCAAQFhAeMgYIABAWEB46BwgAEEcQsAM6CAgAEOoCEI8BOgUIABCRAjoECAAQQzoICC4QsQMQgwE6CAgAELEDEIMBOgsILhCxAxDHARCjAjoICC4QxwEQowI6BAguEEM6CAgAEMkDEJECOgUIABCxAzoICC4QxwEQrwE6BQguELEDOggIABCxAxCRAjoHCAAQsQMQQzoHCAAQsQMQCjoHCAAQyQMQCjoFCAAQkgM6BAgAEAo6CggAELEDEMkDEApKBQgEEgExSgUIBRIBMUoFCAYSATFKBQgIEgExSgUICRIBMUoGCAoSAjUxULCWAVjumgJgvKMCaAJwAngDgAGlA4gBsC-SAQo3LjMxLjQuMC4xmAEAoAEBqgEHZ3dzLXdperABB8gBCMABAQ&sclient=psy-ab&ved=0ahUKEwiTzeaf87DtAhXlTTABHbaaAU4Q4dUDCAw&uact=5).
May the force be with you, David Booth
PS I forgot that you wanted PLS-SEM see: https://www.google.com/search?client=firefox-b-1-d&ei=s1_IX_6LJpDb5gL27LUI&q=PLS+Structual+equation+modeling+in+R&oq=PLS+Structual+equation+modeling+in+R&gs_lcp=CgZwc3ktYWIQAzoHCAAQRxCwAzoECAAQDToGCAAQDRAeSgUICBIBMVDi7QxY56oNYJHDDWgBcAF4AIABiQGIAb4DkgEDMS4zmAEAoAEBqgEHZ3dzLXdpesgBCMABAQ&sclient=psy-ab&ved=0ahUKEwi-saqy87DtAhWQrVkKHXZ2DQEQ4dUDCAw&uact=5
David Eugene Booth Thank you very much for adding a detailed response. I am actually not an expert programmer that's why I have to rely on other's program/packages.
I really appreciate your support.
My concerns are as below:
1. I actually wanted to understand the same model differences for various age group. But, I am struggling how to justify it statistically. For example, is there any statistical issue to use the same path model for various age groups and interpret age specific results.
2. I can't use Lavaan based program as because data doesn't follow normality assumptions. And, the question is still there how to compare two age specific specific which has same constructs and paths.
3. For PLSPM, for general model (whole population). I am unable to compare more than three sub-groups. I am allowed only to compare binary groups.
Belal Edries Thanks for your detailed answer. I am gonna try for the approach you have suggested. Let me see if the results are interpretable and interesting. Thanks
Dear Farrukh Baig
Altough the plspm package provides you a set of wide range of abilities, however, the respecification of the structural model can be a bit time consuming, especially if the model is relatively complicated. Therefore, I would recommend you to use SmartPLS, which is a powerful software that is designed for handling SEM-PLS and it has a nice graphical interface which facilitate the specification or any respecification you might want to include. Nevertheless, it already has a powerful options to run a wide range of analysis types that you want to use, including the one in question of the multi group analysis.
Best Regards,
Belal Edries
Belal Edries I do agree with you that smartPLS is very advanced option for SEM anlaysis. Though, I have used it before and got promising results. But, it has limitations as it is not open source software. Thank you for your suggestion.
Best regards,
Farrukh Baig
If it is a model with a dependent variable and includes indirect effects o rutas (A-B-C) you can evaluate with the free access program JAMOVI because these models are based on SEM and path models, and include conditional effects by category or factor (type ANOVA) that would allow you to estimate the effect of moderate regressions for each of the three age groups.
https://jamovi-amm.github.io/glm.html#factors_coding
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling