Back-projection of regression coefficients onto the original space

This function projects - based on Loadings as provided by reduce_features_FC() - regression coefficients onto the original space (i.e. a square matrix or a volume). Coefficients can either be passed as a list (as created, for example, by FCnetLOO()), as a data.frame, or as a user-defined vector. If a single value for the coefficients is provided, then the function understands that the user wants to retrieve the edges' loadings for a specific coefficient, i.e. (possibly) a network. The function returns a square matrix or a 3-D array whose entries are the linear combination of the coefficients and the edge's loadings. Optional thresholding can be passed to the threshold parameter: an integer value must be specified in order to retrieve the k largest coefficients, in absolute value, only.

backprojectFCnet(coeffs, reduce_features_object, threshold = NULL)

Arguments

coeffs	Coefficients can either be passed as a list (as created, for example, by FCnetLOO()), as a data.frame, or as a user-defined vector. If you fetch a list or a data.frame, then all the coefficients will be backprojected: you may want to select those that are significant following a permutation test instead (see select_coefficients()). If a single value for the coefficients is provided, then the function understands that the user wants to retrieve the edges' loadings for a specific coefficients, i.e. (possibly) a network.
reduce_features_object	An object created by reduce_features_FC(). If present, Loadings may not be provided, and is overwritten.
threshold	Optional. Prune the back-projection matrix by retaining only the threshold larger entries (in absolute value).

Value

A square back-projection matrix or 3-D volume with the original dimensions.