timeLagLasso: Fit a time-lag lasso

In orderedLasso: Ordered Lasso and Time-Lag Sparse Regression

Description

Fit a time-lag lasso model. Builds a regression model with multiple predictors, where an ordered constraint is imposed on each predictor.

Usage

timeLagLasso(x, y, lambda, maxlag, intercept = TRUE, standardize = TRUE,
  beta_pos = NULL, beta_neg = NULL, b0 = NULL, strongly.ordered = TRUE,
  maxiter = 500, inneriter = 100, iter.gg = 100, method = c("Solve.QP",
  "GG"), trace = FALSE, epsilon = 1e-05)

Arguments

x	A matrix of predictors, where the rows are the samples and the columns are the predictors.
y	A vector of observations, where length(y) equals nrow(x).
lambda	Regularization parameter(>0).
maxlag	maximum time-lag specified by user.
intercept	TRUE if there is an intercept in the model. center=FALSE should almost never be used. This option is available for special uses only.
standardize	Standardize the data matrix x.
beta_pos	Optional vector of initialization of the positive part of the coefficients.
beta_neg	Optional vector of initialization of the negative part of the coefficients.
b0	Initialization of the intercept.
strongly.ordered	An option which allows users to order each predictor coefficients non-increasing in absolute value. Details can be seen in the orderedLasso description; default TRUE.
maxiter	Maximum iterations run by time-lag lasso. Initialized to 500.
inneriter	Maximum iterations run by orderedLasso. Initialized to 100.
iter.gg	Maximum iterations run by generalized gradient. Intialized to 100
method	Two options available, Solve.QP and Generalized Gradient
trace	Output option; trace = TRUE gives verbose output.
epsilon	Error tolerance parameter for convergence criterion; default 1e-5.

Details

First transfer the data matrix x to the correct format corresponding to the maxlag specified by user. Then use coordinate descent method by calling orderedLasso to get updates for each predictor.

Value

bp	Estimated coefficients of the positive part
bn	Estimated coefficients of the negative part
beta	Estimated coefficients of predictors, which are equal to bp - bn
b0	Estimated intercept if there is one in the model
fitted	Fitted values of y
maxlag	Maximum time-lag
p	The number of predictors
type	Type of model fit, "gaussian" in this case

Examples

set.seed(3)
n = 50
maxlag = 5
num_rows_needed = n + maxlag + 1
sigma = 4
x = matrix(rnorm(num_rows_needed * 4), nrow = num_rows_needed)
x_new = time_lag_matrix(x, maxlag)
b = c(3,1,1,0,0,
      4,1,0,0,0,
      3,2,1,0,0,
      1,0,0,0,0)
y = x_new %*% b + sigma* rnorm(nrow(x_new))
y = as.vector(y)
y = c(y, rnorm(maxlag + 1))
model1 = timeLagLasso(x = x, y = y, lambda =  1, maxlag = maxlag, 
          method = "Solve.QP", strongly.ordered = TRUE)

Example output

Loading required package: Matrix
build the matrix for timeLagLasso,  6 observations in y are deleted 

orderedLasso documentation built on May 2, 2019, 6:36 a.m.