It can make sense to run separate analyses in a first step to simplify the analysis. This can make it easier for you to detect sources of model misfit (if there are any). In a larger, more complex model, it can be more difficult to identify the causes of model misfit. In the final step, you probably want to look at all variables combined to study relationships between the different constructs/sources/time points.

Determining whether to use one consolidated CFA model or multiple CFA models based on selected variables depends on several factors, including the research question, the complexity of the design, the relationships between variables, and the available statistical resources. Here are some guidelines to consider when making this decision:

In summary, whether to use a single CFA model or multiple CFA models depends on several factors, including the complexity of the research design, the research question, the correlations between variables, the available statistical resources, and the fit of the models. It is important to carefully consider these factors and seek expert guidance to make an informed decision.

I'd recommend three steps:

1) A parallel analysis to decide on how many factors to extract (your most important decision).

2) Given a large enough sample, random division into two roughly equal sub-samples one for exploratory factor analysis (EFA) and one for CFA. Given a relatively small ratio of cases to variables, I'd forgo this division.

3) Either way, I'd begin with an EFA and use it to develop a model for CFA testing. If for reason of a small ratio of cases to variables, both EFA and CFA

are based on the same cases, your CFA may be compromised by capitalization on chance.