I have one multicategorical (grouping variable with 4 groups of participants) and one continuous predictor, continuous mediator and outcome which makes 4 repeated measure. I am interesting in how to set model in path analysis, or do I need HLM or something else? For multicategorical predictor, should I set relations between dummy variables or not and how to include repeated measures in path analysis. It is not longitudinal data, but there is time schedule, so I suppose that it could be set like longitudinal data, but I am not sure. Thanks.
Thank you very much for your answer. How you described, multi-group model is what I need, but in my case it would be MG-LGM, because my outcome is repeated measure. My only concern is about minimum participants per group. Do you know what are recommendations for this minimum?
I use already lavaan for CFA, so I suppose that I will figure out how to set multi-group path model. Thank you again.
Best regards,
Bojana
Hi, Ronny Scherer,
I've just conduct MG-LGM. However, I've no idea to compare slope or intercept across groups. May I have some information in detail ?
Hi,
I have been working with SEM and many decisions are not easy as for ordinary regression.
If you use your groups as MultiGroup analysis, then you will test invariance of effect (what epidemiologists call effect modification or interactions in ordinary regression). If you introduce the categories (excluding one as baseline) you will model it as covariates (mediators or confounding variables).
The issue of repeated measures is more complex. Growth model is fine, but assumes a "model". Of course, you can also include the previous observation as the main cause of the following one, for example, the weight in month 1 is the cause of weight in month 2. But in this case you are not modelling the growth. There are advantages and disadvantages.
I will be glad to hear more about possibilities you have tried.
Cheers
- Badges
- Science topic