spikeSlabGAM fitReturns basic information about prior and model structure, posterior means of inclusion probabilities, term importance and the most probable models found by the SSVS.
# S3 method for spikeSlabGAM
summary(object, threshold = 0.5, ...)an object of class spikeSlabGAM
threshold for inclusion of a model term. Defaults to 0.5.
arguments passed from or to other methods (not used)
an object of class summary.spikeSlabGAM
Two scalar summaries of term importance are included:
P(gamma = 1)The posterior mean of \(P(\gamma = 1)\), i.e. the marginal posterior inclusion probability of the term.
piThe scaled dot products of the posterior mean of the linear predictor \(\eta_i\) associated with the \(i\)-th term and the total linear predictor \(\eta\): \(\pi_i = \eta_i^T \eta/ \eta^T \eta\). They sum to one (but can be negative as well), and (for mutually orthogonal \(\eta_i\)) provide a percentage decomposition of the sum of squares of \(\eta\). See references for details.
The summary also shows the dimensionality of the basis associated with a term.
The top row in the model table shows the relative frequency of the respective model, the bottom row shows cumulative relative frequencies for the models visited most often.
Gu, Chong (2002). Smoothing Spline ANOVA models. Springer. (see chapter 3.6)