Charts and Graphs 35

8. If a decision is made to eliminate points as discussed in step 7, delete the respective

entries in the data cells. The deletions will appear on the chart, and a new trend line
and correlation equation will be displayed, based on the remaining data points.
9. Make nal adjustments to the cosmetics of the chart, fonts, titles, etc. If a large
number of data points are involved, some adjustments in the size of data markers
or in the line width for the trend line may be in order.

Two examples of power law correlation plots are shown in Figure 3.6. One has a rather
good t, whereas the other has a lot of scatter. In the latter case, one should suspect that
either the data are bad or that a power law relation does not t the physical situation.

FIGURE 3.6
36 What Every Engineer Should Know About Excel

3.10 Use of Logarithmic Scales

The data are rst plotted on a linear graph as shown in Figure 3.7a, indicating a decay-
ing exponential or inverse power relation. Logarithmic scales are then selected by double
clicking on each value axis. A FORMAT AXIS window will appear on the right side of the
Excel worksheet. Under AXIS OPTIONS, select the upper and lower bounds for the axis
as well as the major and minor units on the axis scale. Click Logarithmic Scale and the
desired base (10 is default). Remember to repeat for the other value axis. For the y-axis, in
the FORMAT AXIS section labeled “Horizontal axis crosses,” set the “Axis value” eld to
0.1 (the lower edge of the graph), and the result is shown in Figure 3.7b. Next, a trend line is
added by clicking on the chart. Then, click CHART TOOLS/DESIGN/CHART LAYOUTS/
Add Chart Element/Trendline/More Trendline Options. Under Trendline Options, select
Power, and click Display Equation on the chart and Display R-squared value on the chart.
The chart will automatically update with the trend line, the equation, and the R2 value.
The result is shown in Figure 3.7c. Visual inspection indicates that a power relation does
indeed t the data.

(a)

FIGURE 3.7
(Continued)
Charts and Graphs 37

(b)

(c)

FIGURE 3.7 (CONTINUED)

3.11 Correlation with Exponential Functions

The exponential function y = e−0.1x (y = EXP(−0.01x) as an Excel function) is tabulated and is
shown rst as a linear plot in Figure 3.8a with a linear trend line t, which obviously does
not t. Second, a linear plot with an exponential trend line t is shown in Figure 3.8b with
perfect correlation. Third, the function is plotted on a semi-log graph that displays the
38 What Every Engineer Should Know About Excel

FIGURE 3.8

function as a straight line in Figure 3.8c. Again, an exponential trend line is tted with per-
fect correlation. Inspection of the visual display is needed to evaluate the trend line t. For
comparison, the nal two plots of Figure 3.8d and e show ts of second- and third-degree
polynomials. The third-degree polynomial shows a perfect correlation. Polynomials may
frequently be employed to obtain a good t when the functional form is uncertain.

3.12 Use of Different Scatter Graphs for the Same Data

Figure 3.9 shows six scatter plots of a set of hypothetical experimental data displayed in
the upper-left corner of the sheet. Figure 3.9a is a type 1 scatter graph, Figure 3.9b is a type
4 chart, and Figure 3.9c is a type 3 chart—all plotted with linear scales on both axes (see
Section 3.3). Figure 3.9d through f are the same types of plots, but with logarithmic scales on
the axes (see Section 3.10). The graph in Figure 3.9d shows that the data fall approximately
Charts and Graphs 39

FIGURE 3.9

on a straight line so a power law relation might be anticipated. Inserting a trend line and the
correlation equation and value of R2 (see Section 3.9) provides con rmation of such.
Inspecting the data plot in Figure 3.10a, ve data points seem out of place and are, hence,
suspect. Four of those data points are circled in Figure 3.10a and one is circled in Figure
3.10d. Figure 3.10b and c are type 4 and type 3 scatter charts respectively for the data in
Figure 3.10a. Figure 3.10e and f are also type 4 and type 3 scatter charts, but with logarith-
mic scales on the axes. If these points are eliminated as shown in Figure 3.11, a better cor-
relation results. Similar to Figures 3.9 and 3.10, Figure 3.11 provides type 1, type 4 and type
3 scatter charts in Figure 3.11a through c respectively, and type 1, type 4 and type 3 scatter
charts plotted on a logarithmic scale for Figure 3.11d through f respectively.

3.12.1 Observations
The charts in Figure 3.9c and f do not convey much information about the data and do not
give the reader any hint of what might be going on with the experiment. Looking at the
other charts would certainly not give one the impression of a smooth variation of y as a
function of x. The charts in Figure 3.9b and e are better, but those in Figure 3.9a and d give
the best impression of the scatter of data. The chart in Figure 3.9d, because it indicates that
the data are approximately on a straight line in a log–log plot, gives the clue that a power