get_trend_BIC: Choose smoothing parameter using SIC

In halleybrantley/detrendr: Quantile Trend Filtering

Description

get_trend_cv Selects smoothing parameter using Schwarz Information

Usage

get_trend_BIC(y, tau, k, lambdaSeq = exp(seq(0, 14, 1)),
  df_tol = 1e-09, gamma = 1, plot_lambda = FALSE, solver = NULL,
  criteria = "eBIC")

Arguments

y	observed data
tau	quantile levels at which to evaluate trend
k	order of differencing
lambdaSeq	smoothing penalty parameter options to compare
df_tol	tolerance for determining degrees of freedom (Dtheta > df_tol)
gamma	parameter for eBIC
plot_lambda	TRUE/FALSE for plotting lambda by model criteria
solver	LP solver, can be "gurobi", "Rglpk", or "lpSolve"
criteria	criteria to use for lambda selection, must be "eBIC", "SIC", or "valid"

Examples

require(Matrix)
n <- 100
x <- seq(1, n, 1)
y <- sin(x*2*pi/n) + rnorm(n, 0, .4)
lambdaSeq <- exp(seq(-2, 5, 1))
k <- 3
tau <- c(0.05, .2)
trend <- get_trend_BIC(y, tau, k, lambdaSeq, plot_lambda = TRUE)
plot(trend$trend[,1]~x, type="l")  

halleybrantley/detrendr documentation built on May 30, 2019, 12:43 p.m.