MISSING Subcommand (MULTIPLE CORRESPONDENCE command)
In MULTIPLE CORRESPONDENCE, system-missing values, user-defined missing values, and values
less than 1 are treated as missing values. However, in discretizing
a variable, values less than 1 are considered as valid values. The MISSING subcommand allows you to indicate
how to handle missing values for each variable.
PASSIVE. Exclude missing
values on a variable from analysis. This is the default
applicable to all variables, when the MISSING subcommand is omitted or specified without variable names or keywords.
Also, any variable which is not included in the subcommand gets this
specification. Passive treatment of missing values means that, in
optimizing the quantification of a variable, only objects with non-missing
values on the variable are involved and that only the non-missing
values of variables contribute to the solution. Thus, when PASSIVE is specified, missing values do
not affect the analysis. If an object has only missing values, and
for all variables the MISSING option is passive, the object will be handled as a supplementary
object. If on the PRINT subcommand,
correlations are requested and passive treatment of missing values
is specified for a variable, the missing values have to be imputed.
For the correlations of the original variables, missing values on
a variable are imputed with the most frequent category (mode) of the
variable.
ACTIVE. Impute missing values. You can choose to use mode imputation, or to consider objects with missing values on a variable as belonging to the same category and impute missing values with an extra category indicator.
LISTWISE. Exclude cases with missing values on the specified variable(s). The cases used in the analysis are cases without missing values on the variable(s) specified. Also, any variable that is not included in the subcommand gets this specification.
- The
ALLkeyword may be used to indicate all variables. If it is used, it must be the only variable specification. - A mode or extracat imputation is done before listwise deletion.