User Guide

547
Multidimensio
nal Scaling
Transform Values. In certain cases, such as when variables are measured on very
different scales, you may want to standardize values before computing proximities
(not applic
able to binary data). Select a standardization method from the Standardize
drop-down list (if no standardization is required, select
None).
Multidimen
sional Scaling Model
Figure 38-4
Multidimensional Scaling Model dialog box
Correct
estimation of a multidimensional scaling model depends on aspects of the
data and the model itself.
Level of measurement. Allows you to specify the level of your data. Alternatives
are Ord
inal, Interval, or Ratio. If your variables are ordinal, selecting
Untie tied
observations
requests that they be treated as continuous variables, so that ties (equal
values for different cases) are resolved optimally.
Condit
ionality.
Allows you to specify which comparisons are meaningful. Alternatives
are Matrix, Row, or Unconditional.
Dimensions. Allows you to specify the dimensionality of the scaling solution(s). One
solut
ioniscalculatedforeachnumberintherange.Specifyintegersbetween1and6;
a minimum of 1 is allowed only if you select
Euclidean distance as the scaling model.
For a single solution, specify the same number as minimum and maximum.