home *** CD-ROM | disk | FTP | other *** search
- ON ECHO
- !*******************************************************************!
- ! !
- ! Dummy Variables !
- ! (Section 6.4) !
- ! !
- !*******************************************************************!
- !
- ! SORITEC Sampler generates dummy variables both with the TIME
- ! command, as illustrated in the TWOSTAGE example, or the DUMMY
- ! command. DUMMY is used to create a periodic dummy variable
- ! consisting of 1's and 0's. It is used here in an adaption of an
- ! analysis by R. Pindyck and S. Roberts, "Optimal Policies for
- ! Monetary Control" ,Annals of Economic and Social Measurement,
- ! January 1974, as described in R.S. Pindyck and D.L. Rubinfeld,
- ! Econometric Models and Economic Forecasts. New York:
- ! McGraw-Hill, 1976. Several changes have been made to the
- ! original specification to allow the model to be estimated within
- ! the limited workspace of SORITEC Sampler and to accommodate
- ! changes in data since the original study was done. Although
- ! results are different, the techniques employed to estimate the
- ! model are similar.
- !
- ! This example estimates an equation to predict the total volume
- ! of large denomination time deposits > $100,000 in relation to
- ! (1) large denomination time deposits from the previous period
- ! (2) interest rates on 6 month tbills
- ! (3) interest rates on 6 month CD's
- ! (4) personal income
- ! These variables are read in below.
-
- USE 1974M12 1979M12
- READ large_time_deposits interest_tbill interest_CDs personal_income
- 145.5 7.0910 8.820 1209.9
- 147.9 6.5250 7.420 1208.7
- 145.0 5.6740 6.430 1214.3
- 142.0 5.6360 6.330 1219.7
- 137.1 6.0120 6.750 1228.0
- 135.2 5.6490 6.300 1242.1
- 131.5 5.4630 6.250 1272.3
- 128.8 6.4920 7.060 1269.5
- 129.4 6.9400 7.610 1285.6
- 130.4 6.8700 7.890 1295.5
- 131.1 6.3850 7.230 1307.0
- 131.0 5.7510 6.840 1315.2
- 131.1 5.9330 6.560 1321.7
- 128.3 5.2380 5.570 1338.6
- 124.3 5.1440 5.680 1350.8
- 125.3 5.4880 5.870 1356.2
- 124.1 5.2010 5.520 1365.8
- 121.2 5.6000 6.110 1375.0
- 123.3 5.7840 6.310 1380.2
- 124.1 5.5970 5.900 1394.3
- 122.8 5.4160 5.700 1404.8
- 121.2 5.3110 5.580 1413.5
- 119.3 5.0730 5.300 1422.6
- 118.2 4.9440 5.190 1440.5
- 120.0 4.5130 4.760 1452.1
- 118.9 4.7830 5.110 1460.4
- 118.5 4.8960 5.150 1477.1
- 118.9 4.8830 5.160 1493.1
- 117.8 4.7900 5.050 1502.6
- 119.0 5.1930 5.730 1514.1
- 121.8 5.1980 5.640 1526.9
- 124.7 5.3510 5.700 1548.2
- 129.3 5.8100 6.200 1559.7
- 132.2 5.9910 6.450 1575.3
- 136.7 6.4100 6.960 1593.5
- 142.0 6.4330 6.930 1608.9
- 148.0 6.3770 6.950 1625.0
- 150.8 6.6850 7.250 1628.5
- 153.9 6.7400 7.250 1642.9
- 159.5 6.6440 7.180 1663.4
- 163.0 6.7000 7.380 1686.9
- 168.7 7.0190 7.840 1700.9
- 171.1 7.2000 8.240 1720.1
- 174.8 7.4710 8.560 1745.9
- 180.4 7.3630 8.440 1761.6
- 182.6 7.9480 8.960 1777.8
- 184.9 8.4930 9.840 1801.2
- 193.9 9.2040 11.15 1819.5
- 199.3 9.3970 11.28 1843.3
- 200.4 9.5010 11.09 1853.4
- 202.4 9.3490 10.62 1870.2
- 202.7 9.4580 10.47 1891.8
- 198.6 9.4980 10.34 1902.2
- 198.7 9.5310 10.44 1915.8
- 196.5 9.0620 9.980 1931.9
- 198.6 9.1900 10.23 1965.3
- 204.0 9.4500 10.86 1983.1
- 211.1 10.125 12.01 1997.4
- 218.2 11.339 13.83 2016.8
- 222.7 11.856 13.97 2034.2
- 226.3 11.847 13.42 2052.2
- ;
- END
-
-
- ! A set of dummy variables is introduced to explain as much seasonal
- ! behavior as possible. The seasonal variables s1,s2, etc. take the
- ! form of monthly dummy variables multiplied by personal income. Thus,
- ! the first seasonal variable takes the value 1 when the observation
- ! occurs in the month of January, 2 for February, etc., meaning that
- ! there are a total of 12 seasonal dummy variables in the model.
- !
- ! First, reset the USE period.
-
- USE 1975M1 1979M12
-
- !
- ! We'll use two of SORITEC SAMPLER's facilities, the DOT-loop and
- ! direct addressing of subscripts to eliminate some of the
- ! repetition of defining 12 seasonal dummy variables.
- !
- ! First, set up an index variable, j.
- !
-
- SET j=0
-
- !
- ! The way we define the dummy variables generates MISSING values
- ! prior to the first non-zero value (we could avoid this by defining
- ! the USE period from 1974M1). Since we are going to perform
- ! mathematical operations on the data, we must tell SORITEC to
- ! impute MISSING values as zero. We do this by the command:
-
- IMPUTE ZERO
-
- ! Now start the DOT loop.
-
- DOT s1 s2 s3 s4 s5 s6 s7 s8 s9 s10 s11 s12
-
- DUMMY :(j) 1975m1 12
-
- !
- ! Since we must multiply the dummy variables by personal income,
- ! do it while we are still in the loop.
- !
-
- : = : * personal_income
-
- !
- ! Increment the subscript index.
- !
-
- SET j = j + 1
-
- ENDDOT
-
- !
- ! A partial listing of the dummy variables reveals what they look like.
- !
-
- USE 1979m1 1979m6
- PRINT s1 s2 s3 s4 s5 s6
- USE 1975m1 1979m12
-
- !
- ! Since transformations are not permitted as arguments in SORITEC's
- ! estimation commands, they must be performed before estimating the
- ! model. In addition to the 12 seasonal dummies, independent variables
- ! in the model are:
- ! (1) large_time_deposits(-1)
- ! (2) interest_tbill * personal_income
- ! (3) interest_CDs * personal_income
- !
- ! Lags can be specified in estimation commands so we do not need to
- ! transform the first independent variable. Transformations (2) and
- ! (3) can be executed in a DOT-loop.
- !
-
- DOT interest_tbill interest_CDs
-
- !
- ! We'll indicate that the independent variables have been
- ! transformed by prefixing each variable name by pi_.
- !
-
- pi_: = : * personal_income
- ENDDOT
-
- !
- ! Now estimate the equation with the REGRESS command! Similar to
- ! the Pindyck and Roberts study, we shall constrain the fitted
- ! equation through the origin by adding the ORIGIN keyword to the
- ! command line. We'll also refer to the 12 seasonal dummies as a
- ! GROUP so that we don't have to enter the 12 variable names.
- !
-
- !
- ! First define the dummy variables in a GROUP, i.e.,
- !
-
- GROUP dummy_group s1 s2 s3 s4 s5 s6 s7 s8 s9 s10 s11 s12
-
- !
- ! Now permit GROUP expansion by enabling the GROUP option
- !
-
- ON GROUP
-
- !
- ! This is a fairly large problem for SORITEC Sampler's limited
- ! workspace. You can conserve space by disabling the NOMATS
- ! option with the OFF NOMATS command. This suppresses the
- ! storage of the variance-covariance and correlation matrices
- ! associated with the regression.
- !
-
- OFF NOMATS
-
- REGRESS(ORIGIN) large_time_deposits large_time_deposits(-1) &
- pi_interest_tbill pi_interest_CDs &
- dummy_group
-
- !
- ! Turn off the impute flag.
- !
-
- IMPUTE NONE
-
- !
- ! That's it! Note that the seasonal dummies exhibit similar
- ! seasonal trends to those in the original study.
- !
- QUIT
-
-
