6. SIMPLE REGRESSION ANALYSIS › 6.6 Case Studies › 6.6.1 Extended Example I/O Workload Analysis › 6.6.1.3 Output Report

6.6.1.3 Output Report

Simple Regression Analysis produces the plot shown in Figure
6-7.  This plot illustrates some of the problems you may
encounter with historical data. (These types of data behavior
problems are discussed in Section 6.2.)  Although the
historical observations from September 1997 through April
1998 are reasonably well behaved, the six monthly values
recorded from May 1998 through August 1998 illustrate a great
deal of variability.  If the program had executed with a
historical data series ending in May or June 1998, the
resulting trend would have been negative.  Therefore, you
should be hesitant to base an extended forecast on this data
series.

Investigation of the historical data might reveal some
reasons for the variations in their values.  For example, a
week of data might have been lost in May or a new workload
might have been added to the system starting in June.
Univariate Model Forecasting, which is discussed in Chapter
7, allows you to apply the results of this type of
investigation by deleting or adjusting historical data
points.

                                                          EXTENDED EXAMPLE
                                                    CHANNEL 1 SIO COUNT VS MONTH
                                                       SYSID: TSO1   ZONE: 1

                                    PLOT OF CHAN1CNT*DATE    LEGEND: A = 1 OBS, B = 2 OBS, ETC.
                                    PLOT OF PX*DATE          SYMBOL USED IS *
                                    PLOT OF EST*DATE         SYMBOL USED IS +


            15000000 +



                                                                             A     A
            14000000 +



                                                                                                                   +
          C 13000000 +                                                                                        +
          H                                                                                              +
          A                                                                                        +
          N                                                                                   +
                                                                                        +
          1 12000000 +                                                             *
                                                                             *
          S                                                             *
          I                A     A                           *    *
          O                                            *

            11000000 +                           A
          C                                 *    *
          O                            *    A                           A
          U                      *
          N           A    *
          T 10000000 +*                A               A
                                                             A



             9000000 +
                                                                  A



             8000000 +


                     -+----------+----------+----------+----------+----------+----------+----------+----------+----------+
                   30SEP97    30NOV97    31JAN98    31MAR98    31MAY98    31JUL98    30SEP98    30NOV98    31JAN99    31MAR99

                                                                 MONTH ENDING

 Figure 6-7.  Extended Example Simple Regression Forecast Plot