Buy Now
Procedures and Statistics for Analyzing Categorical Data
Using SPSS Categories with SPSS
Statistics
Base gives you a selection of statistical techniques for analyzing high-dimensional or categorical data.
- Categorical regression (CATREG) predicts the values of a nominal, ordinal, or numerical outcome variable from a combination of categorical predictor variables. Optimal scaling techniques are used to quantify variables. Three new regularization methods: Ridge regression, the Lasso, and the Elastic Net, improve prediction accuracy by stabilizing the parameter estimates. Automatic variable selection makes it possible to analyze high-volume datasets—more variables than objects. And by using the numeric scaling level, you can do regularization in regression by using the Lasso or the Elastic Net for your numeric data as well. You can also use CATREG to apply particular Generalized Additive Models (GAM), both for your numeric and categorical data.
- Correspondence analysis (CORRESPONDENCE) enables you to analyze two-way tables that contain some measurement of correspondence between rows and columns, as well as display rows and columns as points in a map.
- Multiple correspondence analysis (MULTIPLE CORRESPONDENCE) is used to analyze multivariate categorical data. It differs from correspondence analysis in that it allows you to use more than two variables in your analysis. With this procedure, all the variables are analyzed at the nominal level (unordered categories).
- Categorical principal components analysis (CATPCA) uses optimal scaling to generalize the principal components analysis procedure so that it can accommodate variables of mixed measurement levels. It is similar to multiple correspondence analysis, except that you are able to specify an analysis level on a variable-by-variable basis.
- Nonlinear canonical correlation analysis (OVERALS) uses optimal scaling to generalize the canonical correlation analysis procedure so that it can accommodate variables of mixed measurement levels. This type of analysis enables you to compare multiple sets of variables to one another in the same graph, after removing the correlation within sets.
- Multidimensional scaling (PROXSCAL) performs multidimensional scaling of one or more matrices with similarities or dissimilarities (proximities). Alternatively, you can compute distances between cases in multivariate data as input to PROXSCAL.
- Preference scaling (PREFSCAL) visually examines relationships between two sets of objects, for example, consumers and products. Preference scaling performs multidimensional unfolding in order to find a map that represents the relationships between these two sets of objects as distances between two sets of points.
To learn more:
Buy Now
