btf: Bayesian trend filtering via Eigen

In btf: Estimates Univariate Function via Bayesian Trend Filtering

Description

Fits Bayesian trend filtering hierarchical model to univariate function. Two conditional priors are available: double exponential or generalized double Pareto.

Usage

btf(y = "vector", x = NULL, k = "int", iter = 10000,
  cond.prior = c("gdp", "dexp"), alpha = NULL, rho = NULL, D = "Matrix",
  m = 1, debug = FALSE)

Arguments

y	response vector
x	inputs corresponding to y observations
k	degree of polynomial fit
iter	number of samples to draw from posterior
cond.prior	choose the conditional prior on f|sigma
alpha	shape parameter for prior on lambda
rho	rate parameter for prior on lambda
D	linear transformation of coefficients inside penalty
m	sample f every mth iteration, default is m=1
debug	boolean telling btf to check for NaNs or not

Author(s)

Edward A. Roualdes

References

R. J. Tibshirani. Adaptive piecewise polynomial estimation via trend filtering. The Annals of Statistics, 42(1):285-323, 2014.

See Also

trendfilter

Examples

# Cubic trend filtering
# from genlasso::trendfilter
## Not run: 
n <- 100
beta0 = numeric(100)
beta0[1:40] <- (1:40-20)^3
beta0[40:50] <- -60*(40:50-50)^2 + 60*100+20^3
beta0[50:70] <- -20*(50:70-50)^2 + 60*100+20^3
beta0[70:100] <- -1/6*(70:100-110)^3 + -1/6*40^3 + 6000
beta0 <- -beta0
beta0 <- (beta0-min(beta0))*10/diff(range(beta0))
y <- beta0 + rnorm(n)
bfit <- btf(y=y, k=3)
plot(bfit, col='grey70')
## End(Not run)

btf documentation built on May 31, 2017, 8:22 p.m.