This chapter is from the book
This chapter is from the book
This chapter is from the book
Issues and Techniques Related to Scaling
Xcelsius 2008 provides for auto-scaling of charts. This relieves you of the burden and drudgery of manually setting a chart scale. Most of the time, auto-scaling works well, but if your living is based on presentations and dashboards, you might want more fine-tuned control than auto-scaling allows.
Consider the following data regarding estimates of manufacturing efficiency:
day production efficiency 7 59% 14 88% 21 91% 28 99%
Depending on real-world circumstances, the data scale that auto-scaling chooses may or may not be appropriate (see Figure 5.30). In this example, the scale reaches 120%. In terms of manufacturing efficiency, 120% is a physically meaningless quantity. Except for reporting or rounding errors and incorrectly calculated estimates, manufacturing efficiency would not exceed 100%.
)
Figure 5.30 Auto-scaling can go well beyond the data extremes.
The point here is that there will be times you will want to take charge of how Xcelsius scales the data in your charts. With the aid of spreadsheet formulas you can design, you may be able to create the scaling behavior you are looking for.
Exploring the Scaling Laboratory
Rather than try to explain the intricacies of the various permutations and combinations of scaling settings, in this section I provide you with a scaling laboratory dashboard (see Figure 5.31 or have a go at it with ch05_ScalingLab.xlf).
)
Figure 5.31 Chart scaling dashboard for which you can adjust the data extremes.
In the scaling lab dashboard, you have the option of specifying how minimum and maximum scales are handled.
This dashboard has two data series, which are displayed in a combination chart. The data used for the chart is displayed in a table (on the right side of Figure 5.31). Notice that two of the data points in this table are shaded. You can adjust the values for the two data points by using the vertical sliders immediately above the data table. The vertical sliders allow you to dynamically adjust values plotted on the chart, so you can see what happens based on the prevailing scaling behavior.
You set the scaling behavior by clicking the various options in the two list boxes near the upper-left portion of the dashboard.
Here is a brief description of the various terms in the Minimum list box:
- Use Minimum Value: This is the minimum value of all the data points displayed in the data table. It includes the values from both series.
- minValue - x%: This is the minimum value reduced by an extension factor. You can adjust this extension factor by using the horizontal slider labeled Extension Factor near the top-right side of the dashboard.
- Min - x% of Delta: This takes the minimum value of all data points and sets the lower limit of the scale to be a set percentage of the difference between the maximum and minimum values of both data series. If all your data is concentrated over a narrow range of values, this type of scaling would be appropriate.
- Fixed Min of x: This hardwires the lower limit of the scale to a fixed number. You have the option of setting this value by using a slider. Once you set it, the value is unchanging until you decide to manually revise it.
- Zero based: This option hardwires the scale’s lower limit to 0.
The Maximum list box options are largely the equivalent of those in the Minimum list box, except that they apply to the scale’s upper limit and tend to add rather than subtract values. In addition, there is no zero-based equivalent for the Maximum list box.
Dealing with Vastly Different Values on the Same Chart
Sometimes you can get caught with having quantities such as 10, 100, and 60,000 in the same chart. If you place these on a linear plot, the small values will virtually disappear. If you are tabulating information such as loss or impairment of an asset and frequency of occurrence, then you definitely don’t want to forgo treating the infrequent but very expensive events in your data analysis.
Figure 5.32 shows government-published data on number of oil pipeline accidents versus barrels lost in the United States during 2006. There is a remarkable level of linearity on the upper limit for the number of accidents.
)
Figure 5.32 A LogLog scale (that is, logarithmic scaling on both the X- and Y-axes) reveals structured relationship over many orders of magnitude.
Setting up logarithmic scaling is rather straightforward. You simply open the Scale subtab of the chart’s Behavior tab and select Logarithmic for both Horizontal and Vertical Axis Scale (see Figure 5.33). You can also experiment with applying logarithmic scaling for only one of the axes.
)
Figure 5.33 Specifying logarithmic scaling on an XY chart.
What happens if you keep both axes linear? The details for the smaller values are almost completely lost because they are too small to be seen (see Figure 5.34).
)
Figure 5.34 The smaller values for barrels lost are too small to discern in linear scaling.