Goodness of fit indices for multiple data sources; this node will accept as input multiple data sources, merge them, and compute a single table of goodness of fit for each dependent variable (identified by a unique name). Use this node to compare the goodness of fit for regression or classification problems, when multiple models were used to compute the predicted responses or classifications. Specify the variable with observed values or classifications as the dependent variable (for each data source on input), and specify the variable with predicted values as the (continuous or categorical) predictor variable (for each data source on input). The results for each data source with identical dependent variable names will be placed in the same table for comparison, and summary tables for each unique dependent variable name will be created. Various goodness of fit indices can be computed depending on whether the dependent variable is continuous or categorical.
General
Element Name
Description
Detail of computed results reported
Detail of computed results; if All results is selected, the analysis will report histograms and scatterplots.
Variable type
Specify the variable type of observed and predicted variables.
Continuous
Element Name
Description
Least squares deviation
Specify whether to calculate the least squares deviation (LSD) goodness-of-fit measure.
Average deviation
Specify whether to calculate the average deviation goodness-of-fit measure.
Relative squared error
Specify whether to calculate the relative squared error goodness-of-fit measure; note that all observed values in this case must be greater than zero.
Relative absolute error
Specify whether to calculate the relative absolute error goodness-of-fit measure; note that all observed values in this case must be greater than zero.
Correlation coefficient
Specify whether to calculate the correlation coefficient goodness-of-fit measure.
Categorical
Element Name
Description
Chi-squared test
Specify whether to calculate the Chi-squared test goodness-of-fit measure.
G-squared test
Specify whether to calculate the G-squared test (maximum likelihood Chi-square) goodness-of-fit measure.
Percent disagreement
Specify whether to calculate the percent disagreement goodness-of-fit measure.
Quadratic loss function
Specify whether to calculate the quadratic loss function goodness-of-fit measure.
Information loss function
Specify whether to calculate the information loss function goodness-of-fit measure.
