Dear Stata experts,
Based on data from an experiment, I would like to test the conceptual model depicted at the bottom of this post. In the model, Mediator A, Mediator B and the outcome are latent variables that are measured with 3-4 items each. The independent variable is an observed categorical variable that represents three experimental manipulations (low vs. middle vs. high) to which participants were exposed at the beginning of the experiment.
After having measured measured Mediator A, participants were exposed to a second manipulation (low vs. middle vs. high). After that, Mediator B and the final outcome variable were measured.
I currently struggle to test the model. My initial idea was to estimate a structural equation model with sem, thereby accounting for Manipulation A by including two dummy variables (with "low" as the reference group). To account for manipulation B, I wanted to do a multiple-group analysis with Manipulation B as the grouping variable. Using the estat ginvariant command, I wanted to explore whether paths across the groups are significantly different. Obviously, this is not the right way as the model is not identified.
I would appreciate any hint on how I could test the model.
Based on data from an experiment, I would like to test the conceptual model depicted at the bottom of this post. In the model, Mediator A, Mediator B and the outcome are latent variables that are measured with 3-4 items each. The independent variable is an observed categorical variable that represents three experimental manipulations (low vs. middle vs. high) to which participants were exposed at the beginning of the experiment.
After having measured measured Mediator A, participants were exposed to a second manipulation (low vs. middle vs. high). After that, Mediator B and the final outcome variable were measured.
I currently struggle to test the model. My initial idea was to estimate a structural equation model with sem, thereby accounting for Manipulation A by including two dummy variables (with "low" as the reference group). To account for manipulation B, I wanted to do a multiple-group analysis with Manipulation B as the grouping variable. Using the estat ginvariant command, I wanted to explore whether paths across the groups are significantly different. Obviously, this is not the right way as the model is not identified.
I would appreciate any hint on how I could test the model.
