CASEWISE Subcommand (LOGISTIC REGRESSION command)

CASEWISE produces a casewise listing of the values of the temporary variables that are created by LOGISTIC REGRESSION.

The following keywords are available for specifying temporary variables (see Fox, 1984). When CASEWISE is specified by itself, the default is to list PRED,PGROUP, RESID, ZRESID, and SRESID. If a list of variable names is given, only those named temporary variables are displayed.

PRED. Predicted probability. For each case, the predicted probability of having the second of the two values of the dichotomous dependent variable.

PGROUP. Predicted group. The group to which a case is assigned based on the predicted probability.

RESID. Difference between observed and predicted probabilities.

DEV. Deviance values. For each case, a log-likelihood-ratio statistic, which measures how well the model fits the case, is computed.

LRESID. Logit residual. Residual divided by the product of PRED and 1–PRED.

SRESID. Studentized residual.

ZRESID. Normalized residual. Residual divided by the square root of the product of PRED and 1–PRED.

LEVER. Leverage value. A measure of the relative influence of each observation on the model’s fit.

COOK. Analog of Cook’s influence statistic.

DFBETA. Difference in beta. The difference in the estimated coefficients for each independent variable if the case is omitted.

The following keyword is available for restricting the cases to be displayed, based on the absolute value of SRESID:

OUTLIER (value). Cases with absolute values of SRESID greater than or equal to the specified value are displayed. If OUTLIER is specified with no value, the default is 2.

Example

LOGISTIC REGRESSION VARIABLES = PROMOTED WITH JOBTIME SEX RACE 
 /CATEGORICAL SEX RACE
 /METHOD ENTER
 /CASEWISE SRESID LEVER DFBETA.

CASEWISE produces a casewise listing of the temporary variables SRESID, LEVER, and DFBETA.
There will be one DFBETA value for each parameter in the model. The continuous variable JOBTIME, the two-level categorical variable SEX, and the constant each require one parameter, while the four-level categorical variable RACE requires three parameters. Thus, six values of DFBETA will be produced for each case.