GAM Specifications - Options Tab

Select the Options tab of the GAM Specifications dialog box to access options to set various parameters that determine the convergence of the (iterative) backfitting estimation procedure.

No intercept

Select the No intercept check box if you do not want to include the intercept term in the generalized additive model. Note that if categorical predictors are selected via the Variables button on the GAM Specifications - Quick tab, the intercept is required.  

Inner model estimation

Use the options in the Inner model estimation group box to set the parameters for the inner iterative estimation loop. The iterative estimation procedure will terminate when the likelihood of the data given the model cannot be improved, relative to the threshold parameters specified in the Threshold field, or when the maximum number of iterations specified in the Maximum number of iterations field has been exceeded.

Outer model estimation

Use the options in the Outer model estimation group box to set the parameters for the outer iterative estimation loop. See Inner model estimation (above) for further details.

Note: Outer and inner iterative estimation loop. Detailed descriptions of how generalized additive models are fit to data can be found in Hastie and Tibshirani (1990), as well as Schimek (2000, p. 300). In general, there are two separate iterative operations involved in the algorithm, which are usually labeled the Outer and Inner iterative estimation loop. The purpose of the outer loop is to maximize the overall fit of the model, by minimizing the overall likelihood of the data given the model (similar to the maximum likelihood estimation procedures as described in, for example, the context of Nonlinear Estimation; however, note that specialized algorithms are used to fit generalized additive models). The purpose of the inner loop is to refine the scatterplot smoother, which in the implementation of generalized additive models in STATISTICA is the cubic splines smoother. The smoothing is performed with respect to the partial residuals; i.e., for every predictor k STATISTICA finds the best weighted cubic spline fit that represents the relationship between variable k and the (partial) residuals computed by removing the effect of all other j predictors (j ≠ k).