Hello again Osman,

Here is a link to an on-line calculator by Daniel Sopor for a priori sample size determination for SEM models:

https://www.danielsoper.com/statcalc/calculator.aspx?id=89

Are you referring to Structural Equation Modeling or Scanning Electron Microscopy? You posted your question in both of these forums.

Hello Osman,

I will presume that you are referring to structural equation modeling in your question. The glib answer is, the more cases you have, the merrier!

The more detailed answer is, that it depends on the number of parameter estimates that are required for the model you wish to evaluate. Usual rule of thumb guidelines would be 10-20 cases per parameter estimate (as a lower bound for total N).

The maximum number of parameters is governed by the number of measured variables you incorporate, and equals the sum of the variables used (because you estimate a variance for each), plus the number of unique covariances among the variables. That means, for example, with five measured variables, there would be a maximum of 5 * (5 + 1) / 2 = 15 parameters that could be estimated in what would be called a "fully saturated" or "just identified" model. It is possible to force fewer parameter estimates, usually by asserting that: (a) some of the covariances will be zero; or (b) some sets of parameters will be identical in value.

Here's a link to a more detailed analysis of the question: Article Erratum: Lower bounds on sample size in structural equation ...

Good luck with your work!

Hello again Osman,

Here is a link to an on-line calculator by Daniel Sopor for a priori sample size determination for SEM models:

https://www.danielsoper.com/statcalc/calculator.aspx?id=89