functionalANOVA: generalized functional analysis of variance

Description Usage Arguments Value References


computes a generalized (weighted) functional ANOVA decomposition of a prediction function, giving the best additive decomposition of the prediction function in terms of squared error.


functionalANOVA(data, vars, n = c(10, 2), model, = function(object, newdata) predict(object, newdata = newdata), = NULL)



a data.frame or data.table which contains the features/covariates on which was learned/estimated.


a character vector which corresponds to a subset of the columns of data.


a numeric vector of length 2, where the first element corresponds to the dimension of the grid constructed for each of the elements of vars and the second element to the number of rows to sample from data.


the first argument to, presumably a model object which can make predictions.

a function whose first two arguments are "object" and "newdata" which returns a numeric vector the same length as the number of rows in newdata. the default value is to call the predict method on the model.

a function with two arguments, design and data, both of which are data.tables which returns a numeric of the same length as the number of rows in design. this is intended for use to use the data to estimate the distribution of the input features, and then to use that estimate to the probability of points in the design grid.


a data.table with columns for a grid of points of the vars, a (set of) column(s) that correspond to the estimated effect of those features/covariates on the prediction function, and a column effect which indicates which subset of the covariates/features each estimate belongs to.


Giles Hooker. Generalized Functional ANOVA Diagnostics for High Dimensional Functions of Dependent Variables, Journal of Computational and Graphical Statistics, Vol. 16, No. 3 (2007), pp. 709-732.

zmjones/fanova documentation built on May 4, 2019, 11:24 p.m.