thetafit function

Nodewise LASSO regressions to fit the precision matrix

Fits the precision matrix by running nodewise LASSO regressions.

thetafit(x, parallel = FALSE, ncores = getOption("mc.cores", NULL), 
         intercept = FALSE, K = 20, l = 5, seed = NULL, verbose = FALSE, 
         registerpar = TRUE, ...)

Arguments

• x: T by p data matrix, where T and p respectively denote the sample size and the number of regressors.
• parallel: if TRUE, use parallel foreach to fit nodewise LASSO regressions. Parallel registered within the function.
• ncores: number of cores used in parallelization
• intercept: whether intercept be fitted (TRUE) or set to zero (FALSE). Default is FALSE.
• K: number of folds of the cv loop. Default set to 20.
• l: the gap used to drop observations round test set data. See tscv.sglfit for more details.
• seed: set a value for seed to control results replication, i.e. set.seed(seed) is used. seed is stored in the output list. Default set to as.numeric(Sys.Date()).
• verbose: if TRUE, prints progress bar. Default set to FALSE.
• registerpar: if TRUE, register parallelization using registerDoParallel. Default set to TRUE.
• ...: Other arguments that can be passed to tscv.sglfit .

Returns

thetafit object.

Details

The function runs tscv.sglfit p times by regressing j-th covariate on all other covariates excluding j-th covariate. The precision matrix is then constructed based on LASSO estimates. Each nodewise LASSO regression tuning parameter is optimized using time series cross-validation. See tscv.sglfit for more details on cross-validation implementation.

Examples

set.seed(1)
x = matrix(rnorm(100 * 20), 100, 20)
thetafit(x = x, parallel = FALSE)

Author(s)

Jonas Striaukas