Transcript Technical Issues

LINREG
linreg(options) depvar start end residuals
# list
where: depvar
start end
default
variables in
residuals
do
list
The dependent variable.
The range to use in the regression. The
is the largest common range of all
the regression.
Series name for the residuals. Omit if you
not want to save the regression residuals.
The list of explanatory variables.
Examples of LIN
lin y
# constant x
lin y 1991:12 2001:8
# constant x
lin y 1991:12 2001:8 resids
# constant y{1 2 5}
lin y / resids
# constant y{1 to 4}
Internal Variables
LINREG creates a number of variables that you can use in subsequent
computations. A partial list of these variables is:
%BETA
The coefficient vector. The first coefficient estimated is
%BETA(1), the second %BETA(2), and so on. For example, in the
output for dlja above, the constant is %BETA(1), the coefficient for
dlja{1} is %BETA(2), and so forth.
%tstats
The vector of t-stats
%NDF
Degrees of freedom.
%NOBS
Number of observations.
%NREG
Number of regressors.
%RSS
Residual sum of squares.
%RSQUARED Centered R2 (i.e, the usual measure of R2)
%SEESQ
Standard error of estimate squared
Correlate
Correlate(options) series start end (saveseries)
where: series The series used to compute the correlations.
Results=
series used to save the correlations
NUMBER= The number of autocorrelations to compute. The
default is the integer value of one-fourth the total number of
observations.
PARTIAL= Series for the partial autocorrelations. If you omit
this option, the PACF will not be calculated.
QSTATS
Use this option if you want the Ljung-Box Qstatistics.
SPAN=
Use with qstats to set the width of the intervals
tested. For example, with quarterly data, you can set span = 4,
to obtain Q(4), Q(8), Q(12), and so forth.
The AIC and the SBC
com sbc = nobs*log(%rss) +
%nreg*log(%nobs)
compute aic = %nobs*log(%rss) + 2*%nreg
display 'AIC' aic 'SBC' sbc
BOXJENK
BOXJENK depvar start end residuals
Options
 AR=number of autoregressive parameters [0]
 MA=number of moving average parameters [0]
 iters= number of iterations
 SAR=number of seasonal autoregressive parameters [0]
 SMA=number of seasonal moving average parameters [0]
 DIFFS=number of regular differencings [0]
 SDIFFS=number of seasonal differencings [0]
 CONSTANT/[NOCONSTANT]
Technical Issues
Constant in the equation

box(constant, ar=||1,4||, ma = 2) y
Negative values of the aic and bic

aic = T ln(%rss) + 2*%nobs
To use the aic and bic, the models must be
estimated over the same sample period.


box(constant, ar=||1,4||, ma = 2) dly 90:1 *
box(constant, ar=1, ma = 2) dly 90:1 *
Technical Issues 2
Did not converge error message
The program cannot find the solution for
the coefficients that minimizes the residual
sum of squares.
 increase iters



iters=50
The model is too complex (too
unnecessary many parameters)
Transforming the series
When to difference?
When to use the log?
Graph the transformed series
Check ACF of the transformed series
The ACF
Label the graph of the autocorrelations
Alter bjident.src
 Write in the essential details
 plot the correlations yourself

ACF of the residuals
Bjident
@BJIDENT series start end
0
Options
DIFF=maximum regular
differencings[0]
SDIFFS=maximum seasonal
differencings[0]
TRANS=[NONE]/LOG/ROOT

R
e
1
.
0
0
0
.
7
5
0
.
5
0
0
.
2
5
0
.
0
0
g
u
l
a
r
0
S
e
a
s
o
n
a
l
Transformation to apply to data
[GRAPH]/NOGRAPH
SPAN=seasonal span
-
0
.
2
5
-
0
.
5
0
-
0
.
7
5
C
P
-
1
.
0
O
A
0
0
5
1
0
1
5
2
0
R
R
T
R
I
S
A
Forecast
forecast(print) number steps start
# equation forecasts
number The number of equations in the system. In univariate forecasting,
number is necessarily equal to 1.
steps
The number of forecasts to create.
start
The starting period of the forecasts.
equation
The name of the previously defined equation.
forecastsThe name of the series in which you want to save the
forecasts. This field is optional.
Example
boxjenk(define=eq1,ar=1,ma=1) y / resids
forecast(print) 1 5 101
# eq1
Forecast -- New
FORECAST equations
# equation forecasts
FROM=starting period of the forecast interval
TO=ending period of the forecast interval
STEPS=number of forecast steps to compute
• FROM and TO set the starting and ending periods for
the forecasts, or
• FROM and STEPS to set the starting date and number
of steps (periods)
PRINT/[NOPRINT]
Seasonality in the Box-Jenkins framework
Seasonal AR coefficients

yt = a1yt-1+a12yt-12 + a13yt-13

yt = a1yt-1+a12yt-12 + a1a12yt-13

(1 - a1L)(1 – a12L12)yt
Seasonal MA Coefficients
Seasonal differencing: = yt – yt-12
Seasonality in the Box-Jenkins framework
Seasonal AR coefficients



yt = a1yt-1+a12yt-12 + a13yt-13
yt = a1yt-1+a12yt-12 + a1a12yt-13
(1 - a1L)(1 – a12L12)yt
Seasonal MA Coefficients
Seasonal differencing:

Dyt = yt – yt-1 versus D12yt = yt – yt-12


NOTE: You do not difference 12 times
dif(sdiffs=1) y / sdy