Deal with outliers.

Deal with other messy data, i.e., missingness.

Select an estimator taking into account the randomness of missing data (MCAR, MAR).

Select an estimator taking into account the distribution of the sample data.

Select the correct estimator for the data matrix in hand, e.g., covariance, the correlation –Pearson, polychoric, polyserial-augmented moment, or asymptotic matrix. Specify the matrix type used.

Caution for Heywood cases occur (correlation greater than 1 or negative variance).

Caution for multi-collinearity.

Caution for non-positive definite matrices or/and other inadmissible solutions.

Caution for other convergence problems.

Avoid convergence problems by carefully selecting start values.

Avoid convergence problems by using larger samples.

If the problem persists test the model with an alternative estimator.

Cross-validate the estimation with an alternative estimator with similar assumptions. Are the results comparable?