MGA and MICOM

[Ingrid Snijders]

Dear all,

I'm working on a research project with data time series. Would anyone like me to help a hand?

First, to be sure how to perform the pre-analysis concerning MGA (time point 1 an 2) and, second, to examine cross lagged relationships using SmartPLS. The data collected concerns initially two HOC's, one as IV and the second als DV/MV and last a DV. The cross lagged relationships concern in between the IV dimensions, and second between IV dimensions (HOC-5 dimensions) and MV (HOC-dimensions) and DV.

Let me know if you could help me out here. Thank you so much.

Ingrid

[jmbecker]

If you want to model cross-lagged effects you need to model both time points in one model. The usual assumption there is that you have repeated measures of the same entities (e.g., respondents/individuals). You then need the data in a wide format, so that each respondent is a row and both time points are in that same observational row.

However, with such data it does not make sense to use the common MGA, because your have dependent samples/groups, but the MGA requires independent samples/groups. You would need to do an MGA that accounts for the dependency of your samples. Such a test is currently not available in SmartPLS.

[Abdullah]

If you want to model cross-lagged effects you need to model both time points in one model. The usual assumption there is that you have repeated measures of the same entities (e.g., respondents/individuals). You then need the data in a wide format, so that each respondent is a row and both time points are in that same observational row.

However, with such data it does not make sense to use the common MGA, because your have dependent samples/groups, but the MGA requires independent samples/groups. You would need to do an MGA that accounts for the dependency of your samples. Such a test is currently not available in SmartPLS.

Thank you for this useful information. I'm, however, not sure how to set up the cross-lagged panel in the SmartPLS3. Do we need to allow for paths between variables in the same time point (e.g., COM_T2 and DEL_T2 should be associated or not?) or perhaps include the autoregression paths only? Hope this makes sense.

[sarahk99]

If you want to model cross-lagged effects you need to model both time points in one model. The usual assumption there is that you have repeated measures of the same entities (e.g., respondents/individuals). You then need the data in a wide format click here, so that each respondent is a row and both time points are in that same observational row.

However, with such data it does not make sense to use the common MGA, because your have dependent samples/groups, but the MGA requires independent samples/groups. You would need to do an MGA that accounts for the dependency of your samples. Such a test is currently not available in SmartPLS.

Thank you for this useful information. I'm, however, not sure how to set up the cross-lagged panel in the SmartPLS3. Do we need to allow for paths between variables in the same time point (e.g., COM_T2 and DEL_T2 should be associated or not?) or perhaps include the autoregression paths only? Hope this makes sense.

You need to check out this tutorial!
http://www.quantpsy.org/pubs/zyphur_all ... press).pdf

[Abdullah]

Thanks for sharing this. But I think this paper has demonstrated the method using Mplus, not Smart-PLS. I think cross-lagged panel analysis is not yet implemented in the software. Let's hope this will be included in future updates