.-
help for ^movsumm^                                     (STB-44: sg85)
.-

Moving summaries
----------------

    ^movsumm^ varname [^if^ exp] [^in^ range] [weight], ^g^en^(^newvar^)^
            ^r^esult^(^#^)^ [ ^w^indow^(^#^) e^nd ^m^id 
            { ^b^inomial | ^ow^eight^(^string^)^ } ^wrap^ ]


Description
-----------

^movsumm^ produces a new variable containing moving summaries of varname
for overlapping windows of specified length. varname will usually (but
not necessarily) be a time series with regularly spaced values.

Possible summaries are those produced by ^summarize^ and saved in
^_result( )^.

It is the user's responsibility to place observations in the appropriate
sort order first.

^fweight^s and ^aweight^s are allowed.


Options
-------

^gen(^newvar^)^ specifies newvar as the name for the new variable. It is
    in fact a required option.

^result(^#^)^ specifies which ^_result( )^ from ^summarize^ is to be
    used. It is in fact a required option. See the table below. Note the
    typo in the Manual for 5.0: ^_result(10)^ contains the 50th
    percentile (median).

    1    number of observations
    2    sum of weight
    3    mean
    4    variance
    5    minimum
    6    maximum
    7    5th percentile
    8    10th percentile
    9    25th percentile
   10    50th percentile (median)
   11    75th percentile
   12    90th percentile
   13    95th percentile
   14    skewness
   15    kurtosis
   16    1st percentile
   17    99th percentile
   18    sum of variable

^window(^#^)^ specifies the length of the window, which should be an
    integer at least 2. The default is 3. By default, results for
    odd-length windows are placed in the middle of the window and
    results for even-length windows are placed at the end of the window.
    The defaults can be overridden by ^end^ and ^mid^.

^end^ forces results to be placed at the end of the window.

^mid^ forces results to be placed in the middle of the window, or in the
    case of windows of even length just after it: in the 2nd of 2, the
    3rd of 4, the 4th of 6, and so on.

^binomial^ specifies that binomial weights are used. For a window of
    length k, these are

        k-1          k-1
           C     /  2      for i = 1 to k.
            i-1

^oweight(^string^)^ specifies the user's own weights for values in the
    window. These should be numbers separated by spaces (not commas).
    The number of weights should equal that specified by ^window( )^.
    Thus ^ow(0.3 0.4 0.3)^ specifies weights of 0.3, 0.4, 0.3.

^binomial^ and ^oweight^ are mutually exclusive. Neither may be combined
    with other weights specified by ^[w =^ ...^]^. Each is treated as
    specifying analytic weights, i.e. they are rescaled to sum to the
    number of observations.

^wrap^ specifies that beginning and end values should be calculated by
    wrapping around. Thus with window length 3 and 12 observations the
    window for observation 1 would be 12, 1, 2 and the window for
    observation 12 would be 11, 12, 1. The justification for this option
    would be that the data are periodic, for example time of day or time
    of year.


Remarks
-------

The use of a ^summarize^ command for each window to do the hard work
will mean that much of the computation time is spent producing results
that are not used. Thus it will be often be faster to use some ad hoc
one-line command.

 . ^gen bin3 = (1/4) * (x[_n-1] + 2 * x + x[_n+1])^

or even

 . ^smooth H x, gen(bin3)^

is much faster than

 . ^movsumm x, gen(bin3) bin w(3) r(3)^

The advantages of ^movsumm^ lie principally in being able to produce a
wide range of moving summaries and in being able to wrap around.

Users with non-periodic data may wish to fill in missing beginning and
end values by their own favourite surgery following ^movsumm^.


Examples
--------

        binomial smoothing with window length 7, wrapped around:

        . ^movsumm temp, w(7) r(3) gen(bin7temp) bin wrap^

        moving median, 10 and 90% percentiles, window length 7:

        . ^movsumm temp, w(7) r(10) gen(medtemp)^

        . ^movsumm temp, w(7) r(8) gen(p10temp)^

        . ^movsumm temp, w(7) r(12) gen(p90temp)^

        totals of monthly sales in previous quarter:

        . ^movsumm monsales, w(4) r(18) gen(qtrsales)^

        but this will be slower than

        . ^gen qtrsales = monsales + monsales[_n-1] +^
                          ^monsales[_n-2] + monsales[_n-3]^


Author
------

       Nicholas J. Cox
       University of Durham, U.K.
       n.j.cox@@durham.ac.uk


Also see
--------

    STB:  STB-44 sg85
 Manual:  [R] summarize, [U] 18.1.6 weight
On-line:  help for @summarize@, @smooth@, @egen@