mdfa_analytic: Main estimation routine: Sets-up the generic optimization...

In wiaidp/MDFA: Signalextraction and Forecasting Based on the Multivariate Direct Filter Approach (MDFA)

Description

Main estimation routine: Sets-up the generic optimization criteria proposed in MDFA-Legacy project (book)

Usage

mdfa_analytic(L, lambda, weight_func, Lag, Gamma, eta, cutoff, i1, i2,
  weight_constraint, lambda_cross, lambda_decay, lambda_smooth, lin_eta,
  shift_constraint, grand_mean, b0_H0, c_eta, weight_structure, white_noise,
  synchronicity, lag_mat, troikaner)

Arguments

L	Filter-length
lambda	Customization parameter: Timeliness is emphasized in the ATS-trilemma if lambda>0
weight_func	DFT-matrix or alternative (for example model-based) estimate: first column is the target variable, additional columns are explanatory variables
Lag	Nowcast (Lag=0), Forecast (Lag<0), Backcast (Lag>0)
Gamma	Generic target specification: typically symmetric Lowpass (trend) or Bandpass (cycle) filters. Highpass and anticipative allpass (forecast) can be specified too
eta	Customization parameter: Smoothness is emphasized in the ATS-trilemma if eta>0
cutoff	Specifies start-frequency in stopband from which Smoothness is emphasized (corresponds typically to the cutoff of the lowpass target). Is not used if eta=0.
i1	Boolean. If T a first-order filter constraint in frequency zero is obtained: amplitude of real-time filter must match weight_constraint (handles integration order one)
i2	Boolean. If T a second-order filter constraint in frequency zero is obtained: time-shift of real-time filter must match target (together with i1 handles integration order two)
weight_constraint	Constraint vector in the case i1==T
lambda_cross	Regularization: cross-sectional term
lambda_decay	Regularization: decay term
lambda_smooth	Regularization: smoothness term
lin_eta	Boolean: impose continuous or discontinuous Smoothness customization
shift_constraint	Constraint vector in the case i2==T
grand_mean	Boolean: if T then a grand-mean parametrization is imposed (default is F)
b0_H0	Regularization: shrinkage target (arbitrary designs can be replicated by imposing strong regularization)
c_eta	Boolean: impose mild/strong smoothness customization (default is F)
weight_structure	Add structure to the optimization criterion (default value is weight_structure<-c(0,0): no structure imposed)
white_noise	Impose a flat DFT (phase information is maintained)
synchronicity	Impose a zero shift across series (amplitude information is maintained)
lag_mat	Matrix for implementing effective lags in a mixed-frequency setting
troikaner	Boolean: if T then degrees oif freedom will be computed (time-consuming computations if K is large)

Value

b Matrix of optimal filter coefficients

trffkt Complex transfer function of optimal multivariate filter

rever Criterion value (corresponds to a sample estimate of the MSE if lambda=eta=0)

degrees_freedom Degrees of freedom (when imposing regularization the degrres of freedom are smaller than L times the number of explanatory series)

Accuracy Accuracy term in decomposition of MSE

Smoothness Smoothness term in decomposition of MSE

Timeliness Timeliness term in decomposition of MSE

MS_error Sample estimate of MSE: consistent estimate in the cases lambda>0 and/or eta>0

freezed_degrees_new The complementary of degrees_freedom

wiaidp/MDFA documentation built on June 26, 2019, 1:07 p.m.