Shareware Grab Bag/Shareware Grab Bag.iso/090/sampler4.arc/REVISE.SAC

Text File | 1985-08-30 | 13.4 KB | 360 lines

on echo
!****************************************************************************!
!                                                                            !
!                   Data Revision and Conversion in SORITEC                  !
!                        (Section 2.6 and Section 6.4)                       !
!                                                                            !
!****************************************************************************!
!
!    Data may be revised or their periodicities may be converted by SORITEC
!    Sampler by a simple set of commands.  In general, data revision
!    requires you to enter a sequence of two commands:
!         (1) Set the USE period to the period that you want the
!             data to be revised, and
!         (2) Update the data explicitly with the REVISE command or
!             implicitly through a transformation expression with
!             the ON REVISE global option enabled.
!    Conversion of periodicities is enabled in SORITEC Sampler by the
!    CONVERT command.  All options associated with this command, which
!    identify how conversion is to be accomplished and the output variable
!    name are entered directly into the command line.  The important thing
!    to remember here is that CONVERT converts the data series into the
!    periodicity associated with the currently active USE command.  You
!    need not specify the periodicity of the input series as SORITEC
!    Sampler can detect it.
!
!    The examples shown here demonstrate the fundamentals of data revision
!    and conversion, but do not illustrate all possible applications of the
!    commands.  Additional examples of data revision are given in the
!    SORITEC Sampler documentation.
!
!****************************************************************************!
!                                                                            !
!                               Data Revision                                !
!                               (Section 2.6)                                !
!                                                                            !
!****************************************************************************!
!
!    This example uses estimates of real Gross National Product (GNP)
!    expressed in billions of 1972 dollars to demonstrate how data revision
!    is done in SORITEC.  Read in the data for two separate periods.  Note
!    that the data are quarterly.
!
 USE  1976q1    1979q4
 READ    gnp72_pre1980
    1285.000000         1293.700000         1301.100000         1313.100000
    1341.300000         1363.300000         1385.800000         1388.400000
    1400.000000         1437.000000         1448.800000         1468.400000
    1472.600000         1469.200000         1486.600000         1489.300000
   ;
 USE   1980q1    1984q4
 READ    gnp72
    1496.400000         1461.400000         1464.200000         1477.900000
    1513.500000         1511.700000         1522.100000         1501.300000
    1483.500000         1480.500000         1477.100000         1478.800000
    1491.000000         1524.800000         1550.200000         1572.700000
    1610.900000         1638.800000         1645.200000         1661.100000
   ;
END

!
!    Copy gnp72 into another variable to compare with revised data
!

gnp72_old = gnp72

!
!    Assume that fourth quarter GNP has been revised from the 3.9 percent
!    growth currently implied by the fourth quarter estimate to 5 percent.
!    This is accomplished in a straight-forward manner with the REVISE
!    command.  First, enter the growth rate into SORITEC Sampler using the
!    SET command.
!

SET growth_rate = 1.05

!
!    Second, change the USE period to the time period over which
!    you want to revise the data.  In this case, it is simply 1984Q4.
!

USE 1984q4

!
!    The only thing to remember in this example is that real growth is
!    expressed in annual terms and when estimating GNP from quarter to
!    quarter, it must first be converted to a quarter to quarter growth
!    rate.  This can be done directly within the REVISE command, i.e.,
!

REVISE  gnp72 = gnp72(-1) * growth_rate ** .25

!
!    Compare the revised and original time series by printing the last
!    four quarters, 1984q1 to 1984q4.
!

USE 1984q1 1984q4
PRINT gnp72_old gnp72

!
!    Data can be revised implicitly in transformations if the REVISE
!    global option is set with the ON REVISE command.  This is
!    demonstrated in an example which spices forecasts of GNP over
!    the four quarters of 1985 to our recently revised GNP data.
!
!    A Business Week article, "Why Factories Aren't Running at Full
!    Stream", Business Week, February 18, 1985, p. 118 estimates
!    annual percentage GNP growth for the four quarters of 1985 will
!    be the following:
!
!                  Quarter          Annual Growth Rate(%)
!                  -------          ---------------------
!                   1985q1                  2.6
!                   1985q2                  2.7
!                   1985q3                  2.5
!                   1985q4                  2.3
!
!    We'll enter these numbers into SORITEC Sampler with the FILL command.
!

USE 1985q1 1985q4
FILL growth 1.026 1.027 1.025 1.023

!
!    Now enable data revision global option with the ON REVISE command.
!    When this option is enabled, any transformation of an existing variable
!    changes only the active observations, as defined by the currently active
!    USE period.  Observations outside the active observation range are
!    retained.  If ON REVISE is not set, observations outside the active
!    USE period are lost.
!

ON REVISE

!
!    Simply specify GNP for each quarter of 1985 as a transformation based
!    upon the previous quarter's estimate to revise the GNP time series to
!    include 1985 values.  Remember to convert annual growth rates to
!    quarterly growth rates in the transformation line.  Note that we
!    must enable the DYNAMIC global option to cause the transformation
!    to substitute lagged variables into the equation as it calculates
!    the forecast values.
!

ON DYNAMIC

gnp72 = gnp72(-1) * growth ** 0.25

OFF DYNAMIC

!
!    Don't worry about the error message.  SORITEC Sampler
!    dynamically extends the series, anyway.
!
!    Print out the revised series for 1984 and 1985.
!

USE 1984q1 1985q4
PRINT gnp72

!
!    Remember to disable the REVISE option after you are satisfied with
!    the results.
!

OFF REVISE

!
!    Splicing data series together is just as easy.  Let's say we want
!    to splice our pre-1980 GNP time series to our GNP time series that
!    runs from 1980q1 to 1985q4.  Using the explicit REVISE command, we
!    simply define the USE period over which the data are to be revised,
!    i.e.,
!

USE 1976q1 1979q4

!
!    and then REVISE the data.
!

REVISE gnp72 = gnp72_pre1980

!
!    Print out the entire series to examine the results.
!

USE 1976q1 1985q4
PRINT gnp72

!
!    If you wanted to use the ON REVISE option to splice the two series
!    together, the command sequence would be:
!

USE 1976q1 1979q4
ON REVISE
gnp72 = gnp72_pre1980
OFF REVISE
cls

!****************************************************************************!
!                                                                            !
!                               Data Conversion                              !
!                                (Section 6.4)                               !
!                                                                            !
!****************************************************************************!
!
!    Periodicity conversion is demonstrated by an example where we wish to
!    examine the relationship between real GNP, the federal budget deficit
!    and total population.  Our data consist of "gnp72", which is a quarterly
!    time series, "federal_deficit", which is monthly, and "annual_population"
!    which is an annual time series.  We will change all data series to
!    quarterly data to enable an equation to be estimated.  We will not
!    estimate the equation in this example, however.  Estimation techniques
!    are demonstrated in the "REGRESS.SAC" command file.
!
!    First, read in budget deficit and population data.
!

 USE  1980m1    1984m12
 READ    federal_deficit
   -4.448000000        -9.285000000        -13.10800000         9.751000000
   -14.03700000         12.36800000        -14.99300000        -6.956000000
    5.640000000        -16.92100000        -8.907000000        -1.965000000
   -12.19200000        -15.62000000        -9.582000000         17.19000000
   -16.17100000         15.36300000        -10.34300000        -5.119000000
    6.335000000        -18.10500000        -10.64200000        -19.46800000
    9.339000000        -14.78000000        -18.25500000         9.704000000
   -18.93000000         6.724000000        -19.83100000        -14.70400000
   -1.708000000        -26.16900000        -24.15800000        -17.93800000
   -9.582000000        -25.33600000        -26.03600000        -3.308000000
   -29.28500000         3.401000000        -21.41200000        -17.47700000
    1.946000000        -25.06900000        -21.59100000        -16.66100000
   -5.515000000        -20.38100000        -28.55500000         11.49300000
   -33.93200000        -2.000000000        -16.41600000        -33.49800000
    16.78500000        -28.78700000        -28.46200000             MISSING
   ;
 USE  1979    1984
 READ    annual_population
    225055.0000         227738.0000         230019.0000         232309.0000
    234496.0000             MISSING
   ;
END

!
!    Just for practice, we'll update the two series through 1984.
!
!    (1) federal_deficit
!

USE 1984m12
REVISE federal_deficit = -15.179

!
!    (2) annual_population
!

USE 1984
REVISE annual_population = 236634

!
!    Now convert monthly federal deficit estimates to annual data.
!
!    First, change the USE period to the periodicity to which you
!    want the data converted, in this case, quarterly.
!

USE 1980q1 1984q4

!
!    Second, convert the data using the CONVERT command.  There are
!    5 different options available for converting data from higher to lower
!    frequencies: SUM the observations; AVERAGE the observations; use the
!    MINimun observation in each period; use the MAXimum observation in each
!    period; or use the LAST observation in each period.  The choice depends
!    upon the way the series is defined and on how the converted time series
!    is to be applied.  In our example, it is appropriate to SUM the
!    observations to quarterly periodicity, since federal deficit estimates
!    are expressed as the total deficit incurred in each month.  Since SUM
!    is the default option, the modifier may be omitted from the command
!    line.
!

CONVERT quarterly_fed_deficit = federal_deficit

!
!    Conversion from annual to quarterly data follows the same
!    procedures except that the modifiers are different.  When
!    converting from a lower to higher periodicity, you can
!    FILL each subperiod in the converted series with the value
!    associated with the entire period, or SHARE the data
!    equally among all subperiods.  In this case, neither modifier
!    seems appropriate because FILL produces the same population
!    estimates for each quarter of the year while SHARE clearly
!    generates incorrect results - since quarterly populations are
!    not simply annual population divided by 4!  We can reasonably
!    good results if we CONVERT using the SHARE modifier and
!    then take a moving sum of length 4 over the periods t-2 to t+1.
!    Since SHARE is the default, it can be omitted from the command
!    line.
!

USE 1979q1 1984q4
CONVERT  temp_population = annual_population

!
!    Now calculate the moving sum over the specified period using
!    the MSUM command.
!
!    Since the period over which the moving sum is to be taken ranges
!    from t-2 to t+1, change the active observation period accordingly.
!
USE 1979q2 1984q3

MSUM quarterly_population temp_population(+1) 4

!
!    We'll extrapolate that value from the annual growth rate
!    in population from 1983 to 1984 after adjusting it to a
!    quarterly growth rate.
!
!    Calculate the growth rate.
!

USE 1984
SET growth_rate = (annual_population/annual_population(-1))**0.25

!
!    Extrapolate the quarterly estimate to 1984q4.
!

USE 1984q4
REVISE quarterly_population = quarterly_population(-1) * growth_rate

!
!    Compare the quarterly population estimates with estimates
!    generated by the CONVERT command when the FILL modifier is
!    used.
!

USE 1980q1 1984q4
CONVERT (FILL) temp_population = annual_population
PRINT temp_population quarterly_population

!
!    Now print out the three quarterly series for the 1980q1 to
!    1984q4 period.
!

PRINT gnp72 quarterly_fed_deficit quarterly_population

!
!    That's it!
!
QUIT