SEM and CFA

Model Complexity or/and number of model parameters estimated

Analyses in which all outcome variables are continuous

Normally distributed data, and there are no

Linear effects existing in data

Existing interactions between data

Estimation method

The lower the reliability of the scores the higher the required sample size

Is it a latent variable models or observed variable model?

Less precise data requires larger samples

Missing data require larger sample sizes

CFA in particular

Low number of indicators for the constructs of interest per factor requires larger samples

Lower number of indicators per factor requires larger samples

Indicators that covary highly with multiple factors require larger samples

If the number of factors is high a larger sample is needed

If covariances between factors are low a larger sample is needed