csr: Estimate the Complete Subset Regressions (CSR) with...

In gabrielrvsc/HDeconometrics: Implementation of several econometric models in high-dimension.

Description

Estimate the Complete Subset Regressions Elliott, Gargano and Timmermann (2013) with the possibility of pre-testing the variables as in Garcia, Medeiros and Vasconcelos (2017) and Medeiros and Vasconcelos(2016).

Usage

csr(x, y, K = min(20, ncol(x)), k = 4, fixed.controls = NULL)

Arguments

x	Matrix of independent variables. Each row is an observation and each column is a variable.
y	Response variable equivalent to the function.
K	Number of variables to be selected after the pre-testing. If K=ncol(x) the pre-testing is redundant.
k	Number of variables in each subset. Must be smaller than K.
fixed.controls	A vector indicatin which variables in x are fixed controls. May be a character vector with the variable names or a numeric vector with the variables position in x.

Details

The Complete Subset Regressions estimates, for a given set of K variables, all possible models with k variables. The number of regressions to be fitted grow very fast with K, therefore, in some cases a pre-testing to select only the most relevant variables may be the only way to make the model computationally feasible.

The pre-testing is activated automatically if the total number of variables in x is bigger than K. If the user chooses K=ncol(x) the procedure is exactly the one presented in Elliott, Gargano and Timmermann (2013).

If the user chooses to include fixed controls in the model they will included in all models and each model will have k+length(fixed.controls) variables.

Value

An object with S3 class csr.

coefficients	Coefficients on each model of the subset.
fitted.values	In-sample fitted values.
residuals	Model residuals.
call	The matched call.

References

Elliott, Graham, Antonio Gargano, and Allan Timmermann. "Complete subset regressions." Journal of Econometrics 177.2 (2013): 357-373.

Garcia, Medeiros and Vasconcelos (2017).

Medeiros, Marcelo C., and Gabriel FR Vasconcelos. "Forecasting macroeconomic variables in data-rich environments." Economics Letters 138 (2016): 50-52.

See Also

predict.csr

Examples

## == This example uses the Brazilian inflation data from
#Garcia, Medeiros and Vasconcelos (2017) == ##
data("BRinf")

## == Data preparation == ##
## == The model is yt = a + Xt-1'b + ut == ##
## == The autorregressive is a fixed control == ##
aux = embed(BRinf,2)
y=aux[,1]
x=aux[,-c(1:ncol(BRinf))]

model=csr(x,y,K=20,k=4,fixed.controls = 1)
plot(y,type="l")
lines(fitted(model),col=2)

gabrielrvsc/HDeconometrics documentation built on April 28, 2020, 7:12 a.m.