What’s New in SAS/STAT 14.1

CALIS Procedure

The METHOD=MLSB option in the PROC CALIS statement computes maximum likelihood estimates, Satorra-Bentler scaled chi-square statistics for model fit, and standard error estimates that are based on a sandwich-type formula proposed by Satorra and Bentler (1994). This estimation method is suitable when you apply the normal-theory-based maximum likelihood estimation to either normal or nonnormal data. You can modify the formula of the Satorra-Bentler scaled statistic by using the SBNTW= option. SBNTW=PRED (the default) uses the estimated covariance matrix to compute the normal-theory weight matrix in the formula; SBNTW=OBS uses the observed covariance matrix.

The INFORMATION= option enables you to select either the expected or observed information matrix when you compute standard error estimates. The default is INFORMATION=OBSERVED when you use full information maximum likelihood (FIML) estimation. All other estimation methods use INFORMATION=EXPECTED as the default.

The STDERR=SBSW option enables you to compute standard errors that are based on a sandwich-type formula proposed by Satorra and Bentler (1994). The METHOD=MLSB option automatically sets the STDERR=SBSW option. All other estimation methods use STDERR=UNADJUSTED as the default.

The DESIGNWIDTH= and DESIGNHEIGHT= options in the PATHDIAGRAM statement enable you to change the default design dimensions of path diagrams. You can accommodate more nodes (variables) in the path diagram by setting a larger design dimension. The SCALE= option in the PATHDIAGRAM statement adjusts the relative node size. The TEXTSIZEMIN= option sets the minimum font size for text in path diagrams. These new options apply only to the GRIP layout algorithm (ARRANGE=GRIP).