vbvs_concurrent: vbvs_concurrent

In jeff-goldsmith/vbvs.concurrent: Variational Bayes Variable Selection for the Functional Linear Concurrent Model

Description

Implements variational bayes variable selection for the linear functional concurrent model

Usage

vbvs_concurrent(formula, id.var = NULL, data = NULL, Kt = 5, Kp = 2,
  v0 = 0.01, v1 = 100, standardized = FALSE, t.min = NULL,
  t.max = NULL)

Arguments

formula	formula for desired regression. should have form y ~ x1 + x2 + ... + x_k | t, where t is the variable that parameterized observed functions
id.var	variable giving subject ID vector
data	optional data frame
Kt	number of spline basis functions for coefficients and FPCs
Kp	number of FPCs to estimate
v0	tuning parameter; normal spike variance
v1	tuning parameter; normal slab variance
standardized	logical; are covariates already standardized?
t.min	minimum value to be evaluated on the time domain (useful if data are sparse and / or irregular). if 'NULL', taken to be minium observed value.
t.max	maximum value to be evaluated on the time domain (useful if data are sparse and / or irregular). if 'NULL', taken to be minium observed value.

Author(s)

Jeff Goldsmith [e-mail]

Examples

## Not run: 
library(dplyr)
library(tidyr)
library(refund.shiny)

## set design elements
set.seed(1)
I = 100
p = 50

## coefficient functions
beta1 = function(t) { sin(2*t*pi) }
beta2 = function(t) { cos(2*t*pi) }
beta3 = function(t) { 1 }

## FPC basis functions
psi1 = function(t) { sin(2*t*pi) }
psi2 = function(t) { cos(2*t*pi) }

## generate subjects, observation times, and FPC scores
time.data = sapply(1:I, function(u) {
  ji = sample(10:15, 1)
  rbind(runif(ji, 0, 1) %>% sort, 
        rep(u, ji),
        rep(rnorm(1, 0, 3), ji),
        rep(rnorm(1, 0, 1), ji))
}) %>% unlist() %>% matrix(ncol = 4, byrow = TRUE)
colnames(time.data) = c("time", "subj", "c_i1", "c_i2")
time.data = as.data.frame(time.data)

## generate predictor data
predictor.data = matrix(rnorm(dim(time.data)[1] * p), dim(time.data)[1], p)
colnames(predictor.data) = paste0("Cov_", 1:p)

## combine and generate responses
concurrent.data = cbind(time.data, predictor.data)
concurrent.data = 
  mutate(concurrent.data,
         Y = Cov_1 * beta1(time) +            ## fixed effects
             Cov_2 * beta2(time) + 
             Cov_3 * beta3(time) + 
             c_i1 * psi1(time) +              ## pca effects
             c_i2 * psi2(time) + 
             rnorm(dim(concurrent.data)[1]))  ## measurement error

## fit model
pred.list = paste("Cov", 1:p, sep = "_")
formula = as.formula( paste("Y ~", paste(pred.list, collapse = "+"), "| time") )
fit.vb = vb_concurrent(formula, id.var = "subj", data = concurrent.data, standardized = TRUE,
                           t.min = 0, t.max = 1, Kp = 3)
fit.vbvs = vbvs_concurrent(formula, id.var = "subj", data = concurrent.data, standardized = TRUE,
                           t.min = 0, t.max = 1, Kp = 3)

## interactive plot for mean model
plot_shiny(fit.vb)
plot_shiny(fit.vbvs)

## interactive plot for fpca expansion of residuals
plot_shiny(fit.vbvs$fpca.obj)

## End(Not run)

jeff-goldsmith/vbvs.concurrent documentation built on Sept. 17, 2019, 2:26 p.m.