ANALYSIS Subcommand (PRINCALS command)
ANALYSIS
specifies the variables to be used in the computations and the optimal
scaling level used by PRINCALS
to quantify each variable or variable list.
- If
ANALYSIS
is not specified, an ordinal level of optimal scaling is assumed for all variables. - The specification on
ANALYSIS
is a variable list and an optional keyword in parentheses to indicate the optimal scaling level. - The variables on the variable list must also be specified
on the
VARIABLES
subcommand. - Variables listed on the
VARIABLES
subcommand but not on theANALYSIS
subcommand can still be used to label object scores on thePLOT
subcommand.
The following keywords can be specified to indicate the optimal scaling level:
MNOM. Multiple nominal. The quantifications can be different for each dimension. When all
variables are multiple nominal, PRINCALS
gives the same results as HOMALS
.
SNOM. Single nominal. PRINCALS
gives only one quantification
for each category. Objects in the same category (cases with the same
value on a variable) obtain the same quantification. When DIMENSION
=1 and all variables are SNOM
, this solution is the same as that
of the first HOMALS
dimension.
ORDI. Ordinal. This is the default for variables listed without optimal scaling
levels and for all variables if the ANALYSIS
subcommand is not used. The order of the categories of the observed
variable is preserved in the quantified variable.
NUME. Numerical. This is the interval or ratio level of optimal scaling. PRINCALS
assumes that the observed variable
already has numerical values for its categories. When all variables
are at the numerical level, the PRINCALS
analysis is analogous to classical principal components analysis.
These keywords can apply to a variable list as
well as to a single variable. Thus, the default ORDI
is not applied to a variable without a keyword if
a subsequent variable on the list has a keyword.