*
* Example from page 809
*
all 500
*
* Generate the noise series and the driftless random walk. Note that this will
* produce a different result each time you run it (and won't match the numbers in
* the text). If you want to reproduce the same set of "random" numbers, put a
* SEED instruction near the top of your program.
*
set u = %ran(1.0)
set(first=0.0) ynd = ynd{1}+u
*
* The graphs like those in the text are created using the procedure BJIDENT (BJ
* is short for Box-Jenkins). For the purposes of this, all you need is
*
* @BJIDENT(NUMBER=number of correlations)  series
*
* The confidence bands for the "ynd" series spread out with increasing lag,
* unlike the graphs shown in the text. This is because they use a different null
* hypothesis: those in the text are based upon the null that all correlations are
* zero, while those produced by BJIDENT has a sequence of nulls, where the lag
* L+1 is tested for 0 allowing correlations at lags 1 to L to be non-zero. When
* this is applied to the "u" series, the result is almost identical to assuming
* all are zero, since they're close to zero anyway.
*
@bjident(number=30) u
@bjident(number=33) ynd
*
* With the option DIFFS=1, BJIDENT does the autocorrelation graph of the series
* (0 Differences) and its first difference. When applied to ynd, this will give
* two graphs: the 0 difference will match exactly with the one above for ynd, and
* the 1 difference will be very close to the one above for u. (The only
* difference is that the latter is done with one less data point).
*
@bjident(number=33,diffs=1) ynd