Hello Stefan,

Yes, N = 37 is entirely too few cases.

In general, simulation is the preferred way to gain confidence in what might be plausible lower bounds for sample size in SEM. Part of the reason for this is that models having identical numbers of observed and latent variables, respectively, can still differ in complexity. The Mplus package has been written about frequently for this application. Here's a good resource for pursuing this path:Article Sample Size Requirements for Structural Equation Models: An ...

Here's an alternate bit of advice, for those who'd like to use R: https://stats.stackexchange.com/questions/156726/how-to-run-simulation-in-r-to-assess-sample-size-and-power-for-sem

If that's too daunting, here's more of a cookie-cutter approach, which takes into account minimum ES (as a correlation), target power, and counts of observed and latent variables: https://www.danielsoper.com/statcalc/calculator.aspx?id=89

Good luck with your work.

Stefan Marian

These links might help:

1. https://rdrr.io/cran/semPower/man/semPower.aPriori.html

2. http://www.quantpsy.org/rmsea/rmsea.htm

3. https://psycnet.apa.org/record/2021-97628-001

4. https://github.com/moshagen/semPower

Stefan Marian A few additional comments:

1) your parameter count does not seem correct for the proposed model. You would also typically estimate residual variances for the first-order factors. Those latent residual variances seem to be missing from your free parameter count (q).

2) Your semPower calculation formula/syntax/code in the form in which you used it is limited to studying the behavior of/sample size requirement for the RMSEA fit index. It only tells you about the "sufficient" sample size to reject the null hypothesis of "close fit" using RMSEA (H0: RMSEA