Custom Test allows you to test a custom hypothesis. Refer to in Standard Least Squares Report and Options for details on custom tests.
Contrast allows you to test for differences in levels within a variable. If a contrast involves a covariate, you can specify the value of the covariate at which to test the contrast.
In the Crab Satellite example, suppose you want to test whether female crabs with good spines attracted a different number of male crabs (satellites) than female crabs with worn or broken spines. See for how to run the example. In the report window, selecting Contrast brings up the following window:
Choose spine, the variable of interest, and click Go.
To compare the crabs with good spines to crabs with worn or broken spines, click the + button beside Both Good and the - button beside both One Worn/Broken and Both Worn/Broken.
Since the Prob>Chisq, 0.8242, is much greater than 0.05, we can not conclude that there is a difference in satellite attraction based on spine condition.
Inverse Prediction is used to predict an X value, given specific values for Y and the other X variables. This can be used to predict continuous variables only. For more details about Inverse Prediction, see in Standard Least Squares Report and Options.
Covariance of Estimates produces a covariance matrix for all the effects in a model. The estimated covariance matrix of the parameter estimator is given by
Σ = -H-1
where H is the Hessian (or second derivative) matrix evaluated using the parameter estimates on the last iteration. Note that the dispersion parameter, whether estimated or specified, is incorporated into H. Rows and columns corresponding to aliased parameters are not included in Σ.
Correlation of Estimates produces a correlation matrix for all the effects in a model. The correlation matrix is the normalized covariance matrix. That is, if σij is an element of Σ, then the corresponding element of the correlation matrix is σij/σiσj, where
Profiler brings up the Profiler for examining prediction traces for each X variable. Details on the profiler are found in in Standard Least Squares Report and Options.
Contour Profiler brings up an interactive contour profiler. Details are found in the Profilers book.
Surface Profiler brings up a 3-D surface profiler. Details of Surface Plots are found in the Profilers book.
Diagnostic Plots is a submenu containing commands that allow you to plot combinations of residuals, predicted values, and actual values to search for outliers and determine the adequacy of your model. Deviance is discussed above in . The following plots are available:
Note: Regression Plot is available only when there is one continuous predictor and no more than one categorical predictor.
Note: Linear Predictor Plot is a plot of responses transformed by the inverse link function. This plot is available only when there is one continuous predictor and no more than one categorical predictor.
Model Dialog shows the completed launch window for the current analysis.
Effect Summary shows the interactive Effect Summary report that allows you to add or remove effects from the model. See .
Save Columns is a submenu that lets you save certain quantities as new columns in the data table. Formulas for residuals are shown in Residual Formulas.
where (yi – μi) is the raw residual, sign(yi – μi) is 1 if (yi – μi) is positive and -1 if (yi – μi) is negative, di is the contribution to the total deviance from observation i, φ is the dispersion parameter, V(μi) is the variance function, and hi is the ith diagonal element of the matrix We(1/2)X(X'WeX)-1X'We(1/2), where We is the weight matrix used in computing the expected information matrix. For additional information regarding residuals and generalized linear models, see “The GENMOD Procedure” in the SAS/STAT User Guide documentation.