*
* @ROLLREG( options ) depvar first last
* # list of regressors
*
* This procedure does rolling OLS regressions in one of three modes.
* Given an initial range of observations, ADD adds observations (one at
* a time) to the end of the sample, DROP drops them from the beginning
* of the sample, and MOVE simultaneously adds and drops so that the
* number of observations in each regression is constant.
*
* Parameters:
* depvar is the name of the LHS variable
* first last (optional) maximum range for the regression
* defaults to maximum range for which all variables
* are defined
* Supplementary card:
* list of series lists the RHS variables for the regression
*
* Options:
* ADD= starting with a regression from first to ADD,
* adds one observations at a time until it does a
* regression from first to last
* DROP= starting with a regression from first to last,
* drops one observations at a time until it does a
* regression from DROP to last
* MOVE= starting with a regression from first to first+MOVE-1
* adds and drops one observations at a time until it
* does a regression from last-move+1 to last
* (i.e. there are MOVE observations in each regression)
*
* Note that exactly one of these three options must be used.
*
* e.g. @rollreg(drop=1989:4) gnp
* @rollreg(add=1971:4) gnp
* @rollreg(move=20) gnp
*
* ROBUST/[NOROBUST] uses the ROBUSTERRORS option to calculate the
* standard errors.
* LAGS = [0] number of lags to use with the ROBUST option.
* This does **not** vary with the degrees of freedom.
* LWINDOW=[NEWEY]/BARTLETT/DAMPED/PARZEN/QUADRATIC
* chooses lag window type. Note that the default for
* ROLLREG is the Newey-West, rather than the truncated
* window.
* DAMP = [1.0] shape of the spectral window to use with the
* LWINDOW=DAMPED option.
*
* Output Options:
*
* COHISTORY = VECT[SERIES] for the coefficients
* SEHISTORY = VECT[SERIES] for the standard errors of the coefficients
* SIGHISTORY = SERIES for the regression standard errors
* (These are all similar to the options for the RLS instruction)
*
* FHISTORY = SERIES with regression F's
*
* PRINT]/[NOPRINT] prints the coefficient estimates
* WINDOW = title for print window
* GRAPH/[NOGRAPH] graphs the coefficients with two standard error bands.
*
* Revision Schedule:
* October 1992 Written by Simon van Norden
* August 1993 Modified for RATS v4 by Jeff Gable
* June 1995 Robust Error Bug Fix by Rob Vigfusson
* March 2009 Minor change to description by Tom Doan, Estima
* June 2010 Major reprogramming by Tom Doan, Estima to use newer
* features and make more compatible with syntax of the
* RLS instruction.
* April 2011 Add FHISTORY option. Tom Doan, Estima.
*
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* ROLLREG.SRC
* Copyright 1993 by the Bank of Canada
* Programmed for RATS by Robert Amano, Jeff Gable and Simon van Norden
* In exchange for access to these procedures, users are requested to
* acknowledge the use of the Bank of Canada RATS procedures in
* all published work.
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
*
procedure rollreg y first last
type series y
type integer first last
option integer add 0
option integer drop 0
option integer move 0
option switch robust 0
option choice lwindow 3 newey bartlett damped parzen quadratic flat
option integer lags 0
option real damp 1.0
option switch print 0
option string window
option switch graph 0
option vect[series] *cohistory
option vect[series] *sehistory
option series *sighistory
option series *fhistory
*
local integer fobs lobs cut k fbeta lbeta
local equation eqn
local vect[series] coeff
local vect[series] se
local series upper lower
*
* Define the system to be estimated and initialize
*
* 06/2010 Switch to use EQUATION
*
equation eqn y
*
* 06/2010 Move diagnostic here
*
if (add<>0)+(drop<>0)+(move<>0) != 1 {
disp "###ROLLREG: Must use one and only one of ADD, DROP and MOVE options."
return
}
*
* Get default range
*
inquire(equation=eqn) fobs<