Structural Equation Modeling Results - Residuals Tab

Select the Residuals tab of the Structural Equation Modeling Results dialog box to access the options described here.

Standardized residuals. Click the Standardized residuals button to create a spreadsheet of standardized residuals. Standardized residuals represent the difference between the input matrix and the reproduced matrix, standardized in the correlation metric. We express the residuals in standardized form to eliminate the effects of the scale of the variable on model residuals.

Normalized residuals. Click the Normalized residuals button to create a spreadsheet of normalized residuals. Normalized residuals are divided by an estimate of their variance, to provide a measure that, in large samples, is approximately normally distributed when the model fits perfectly in the population. These residuals can be interpreted (roughly) like any standardized normal statistic.  Keep in mind that, when evaluating a large matrix of residuals, you need to correct for the fact that you are examining, post hoc, a large number of coefficients. For example, you expect approximately 5% of these to be "significant" by chance at the 5% level, so a matrix with 45 residuals would have, typically, 2 or 3 "significant" values by chance.

Normal probability plot. Click the Normal probability plot button to create a normal probability plot of the residuals. You can use the plot to detect departures from normality, perfect fit, or both. If the proposed structural model fits perfectly in the population, and if the population distribution is multivariate normal, the residuals will be approximately normally distributed.

Input matrix. Click the Input matrix button to display the matrix actually analyzed by the structural modeling iterative routine, rather than the data matrix. So, for example, if you input Raw Data, but select Moments as your option under the Data to Analyze group of the Analysis Parameters dialog box, you will see an augmented moment matrix here.

Reproduced matrix. Click the Reproduced matrix button to display the matrix calculated from the model, using the final parameter estimates as the numerical values.