Databases Reference
In-Depth Information
5.9
CREATING A WINDOWING TOOL
Thus far the examples have centered around making a determination if a single condition is met,
and then taking actions based on the outcome of a single condition. More often than not, however,
a result is desired that falls between
two
conditions, creating a “window” in which the desirable
results will reside. Suppose, for example, that the ideal value one is looking for is 100, but an
acceptable range of values could be
10% of the ideal value. In such an instance, a desired value
of (
x
) will fall between 90 and 110 (90
±
110) as illustrated in Figure 5.22. It is worth pointing
out that the window in Figure 5.22 could be looked at in another way. The shaded region could
also be viewed as a region of
exclusion
, in which case a desired value of (x) would be less than
90 or greater than 110 (
x
< 90;
x
> 110).
Although such analyses are useful, it is their rigidity that limits their usefulness. Creating a
rigid boundary leads to all sorts of problems when automatically analyzing data. The difficulties
begin when the value sought is not found within the specified window, yet a value exists just outside
the boundary of the windows range. In the foregoing example, this would occur if a value existed
at 89 or 111. Because these values fall outside the windows range, they would not be chosen.
Invariably, what will happen when this occurs is that a user will complain that even though the
value is outside the defined specification, they would have “let it slip by” because it was “close
enough.” It is easy to see how such analyses can lead to problems when scientists feel that the tool
they are utilizing to analyze their data is excluding potential values of importance to them.
The way around this problem is to not utilize a single window but layered multiple windows,
much like a Venn diagram that displays overlapping choices. Usually, this will consist of three
windows. The first window is inclusive and contains a range around the optimal value that would
be considered valid without question. The second window is also inclusive but encompasses a larger
range around the optimal value. If the value sought is not found within the first window, the window
can then be expanded out to the values of the second window. The second window contains values
that could be termed acceptable but not optimal. The third window is exclusive and contains all
values outside the second window. The third window encompasses the space of unreported values,
meaning, it contains results that are absurd.
Looking at Figure 5.23, it is readily apparent that this method of windowing segregates the possible
regions for a result to exist into three regions of space that can be described as (n
l
≤
x
≤
≤
x
≤
n
u
) {Normal
Range}, (a
l
≤
x
≤
a
u
) {Acceptable Range}, and (x>a
u
; x<a
l
) {The Exclusion Region}; where
x
=
result sought, n
l
=
normal lower range limit, n
u
=
normal upper range limit, a
l
=
acceptable
lower range limit, a
u
=
acceptable upper range limit.
85
90
95 100
Example window
105
110
115
FIGURE 5.22
Sample window interval.
