I am planning to analyze my data with a custom build model in the PROCESS Macro (Hayes). I know that you can calculate model fit indizes with for example AMOS and report them in your thesis. Is is possible to calculate something like that with PROCESS? And if not is it necessary to report them or not? What's the best way to handle this? I am thankful for every input and all tipps you give me. Thanks in advance.
To my knowledge, PROCESS will not give you model fit statistics such as a chi-square test of model fit. Those statistics would only be needed if your mediation (path) model as a whole is over-identified (non-saturated). For over-identified models, I would recommend estimating the model as a simultaneous-equation model within the framework of path analysis using a program for structural equation modeling such as AMOS, Mplus, or the lavaan package in R. Those programs will provide you with fit statistics in addition to all relevant parameter estimates and the direct, indirect (mediated), and total effects.
For saturated (just identified) path models, fit statistics such as the chi-square test are not relevant as saturated models perfectly reproduce the observed covariance structure by definition.
Christian Geiser thank you very much for your answer. In the two added pictures, you can see the two proposed models. I am not sure whether they are saturated or non-saturated. Depending on whether there are gender or grade-level differences, these variables would possibly be added as control variables into the calculation.
The first model is non-saturated because there is no direct path between M1 and M3. I believe the second one is saturated.
Thank you very much. I will have a look at the programs you recommended for non-saturated models.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Social Media