bic_kmean_est: Estimate a partially functional linear regression model with...

In wangwustat/fdagroup: Latent Group Detection in Partially Functional Linear Regression Models

Description

This is the proposed estimator, which is developed based on the functional principal component analysis. The algorithm is a variant of the K-means clustering algorithm. The number of groups and the number of principal components are determined by the BIC criterion.

Usage

bic_kmean_est(data_list, num_group = 2, num_pca = 5,
  est_fix_eff = TRUE, max_iter = 100, loc_search = FALSE,
  group_index = NULL)

Arguments

data_list	A list of data. Several elements must be present in the list. The reponse y, the functional covariate x_recv, the scalar covariates z, and an index matrix index. The functional covariate x_recv must be generated from the fda package by, e.g., spline smoothing. The scalar covariates z is a matrix. The index matrix index is a data.frame recording the structure of the data. The first column of index is the family number, the second column is the within family index. The column names of index must be ind_b and ind_w.
num_group	A vector of candidate number of groups.
num_pca	A vector of candidate number of principal components.
est_fix_eff	A logical value. If TRUE, then the fixed effects are estimated. Otherwise, the fixed effects are not estimated
max_iter	The maximum number of iteration. Default to be 100.
loc_search	A logical value. If TRUE, conduct a local search to decrease the objective function.
group_index	An initial value of the group membership, optional.

wangwustat/fdagroup documentation built on Dec. 5, 2019, 12:51 a.m.