In Category Axis Tricks for Line and Area Charts – 1 I extended this technique to show how to format parts of a line chart in distinct colors:

A reader asked how to show two years of data on two separate lines and show a trendline for each.

Line Chart Approach

Let’s make up two year’s worth of data:

Here’s the data in a two-series timeline chart:

To add a trendline to a series, right click on it and select Add Trendline. Here’s the chart with a trendline for each series.

These trendlines are too long, each extending across the entire date range. The trendline dialog allows us to extend a trendline beyond the data in the forward or backward direction, but it doesn’t allow us to shorten a trendline (the forward and backward text boxes only accept positive values or zeros).

Well, we can always roll our own trendlines. Here is an updated data range:

The slopes and intercepts are calculated as follows:

B27: =SLOPE(B2:B13,A2:A13)

B28: =INTERCEPT(B2:B13,A2:A13)

C27: =SLOPE(C14:C25,A14:A25)

C28: =INTERCEPT(C14:C25,A14:A25)

Column D has trendline data for 2008 with this formula in D2 and filled down to D13:

=$B$28+$B$27*A2

Column E has trendline data for 2009 with this formula in E14 and filled down to E25:

=$C$28+$C$27*A14

The data in these two columns is added to the chart, and a little formatting gives us two distinct timelines with distinct trendlines:

That’s not too bad, but is there any way to use only two series and the native chart trendline feature?

XY Chart Approach

Why not just make an XY chart? Looks the same as the line chart with the homemade trendlines.

But wait, on closer inspection, the date axis looks wrong. The line chart has a date label for every second month, but the XY chart’s axis shows August and October 2008 instead of September and November 2008.

Line charts allow us to adjust the scale to put a label at the first of each month, regardless of the number of days in the previous month. In the line chart, we can set our min and max as 1-Jan-2008 and 1-Jan-2010, and the major and minor tick spacing as 2 months and 1 month (and don’t forget the base unit should be days).

In the XY chart, we can set the same min and max, 1-Jan-2008 and 1-Jan-2010, but we have to use the inconvenient values of 60.917 and 30.4585 days as our major and minor tick spacing (731 days divided by 12 major and 24 minor divisions). This causes some ticks to occur before the start of a month (31-Aug-2008 and 31-Oct-2008 in the chart below), and some after the first of the month (02-Mar-2009, 02-May-2009, and 02-Jul-2009).

The line chart is needed if we want “nice” date scale axis labeling.

XY-Line Combination Approach

Let’s start with a line chart using only the 2008 data, to give us our preferred date axis format.

Copy the 2009 data (select A14:A25, hold CTRL and select C14:C25, A1, and C1, then press CTRL+C), select the chart, and use Paste Special to add the data as a new series. Excel adds a line chart series, using the same X values (dates) as the first series, even though we copied and pasted the X values as well as the Y values.

We can convert the new series to an XY type (right click the series, choose Chart Type, and pick the style we want).

If Excel moves the XY series to the secondary axis as in the chart above, move it back to the primary axis.

Now we can add a trendline to each series, and it extends only as far as the series itself.

Summary

As usual, Excel provides multiple ways to accomplish our goal. We could not do this directly, but we have our choice of at least two viable workarounds: Line chart with manually calculated trendlines, and Line-XY combination chart with built-in trendlines. What’s more tedious, calculating your own trendlines, or creating a combination chart?

Peltier Technical Services, Inc., Copyright © 2009.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

This simple approach uses line charts, and the up-down bars and high-low lines that can be added to line charts. I’ve described these embellishments in Stock Charts and Other Line Chart Tricks. When two or more line chart series are present in a chart, the up-down bars draw vertical bars from the first line series to the last, and the high-low lines are drawn behind the up-down bars (if present) from the lowest to highest line series value. The formatting of up bars (the last series is greater than the first) is distinct from that of down bars (the last series is less than the first).

A stock chart is simply a line chart that uses these embellishments. When data is entered in Open-High-Low-Close order, the up-down bars and high-low lines are used to create candlestick-style stock charts in Excel. Excel 2003 also accepts data in Open-Low-High-Close order, while Excel 2007 insists upon OHLC. Excel 2003 allows the price data to be arranged in columns (as such data is usually provided) or in rows. Excel 2007 insists on price data in columns only. If you start with a line chart and add these graphic features, the arrangement of data is ignored.

Excel 2003 and Earlier

We’ll start with this set of calculations, based on three samples of a larger population that has mean of 25 and standard deviation of 10. The First and Third Quartiles are listed first and last so the up-down bars represent the inner quartiles of the box plot. The Min and Max will be used by the high-low lines, and the intermediate Mean and Median have no effect on the added features.

Here is the line chart produced from the data. (You could instead start with a candlestick stock chart using First Quartile, Min, Max, and Third Quartile data, add the Median and Mean data, and rearrange the series, but I think the way I’ve done it is easier and more straightforward.)

In Excel 2003, the Format Series dialog with any of the lines selected offer Up-Down Bars and High-Low Lines on the Options tab.

Hide all of the series lines, and all markers except for Median and Mean, and we have a serviceable box plot. Here the “Dash” marker is used for median.

If desired, the Min and Max series can also be formatted with a dash marker to provide end caps for the whiskers (high-low lines).

The dash marker is a horizontal line, with a minimum thickness of 2 pixels. The markers above are 14 points wide and still 2 pixels high. At 15 points, the markers expand to 3 pixels. At the size required to match the width of the up-down bars, the dashes are ridiculously thick (and even the two pixel thickness above is too thick).

Particularly ridiculous when used for the end caps too.

Horizontal median lines are relatively easy to obtain. Change the chart type of the median series from Line to XY, and if necessary format it to appear on the primary axes with the line series. Remove its marker, and apply positive and negative horizontal error bars with a value of 0.2.

The default gap width for the up-down bars is 150% of a bar width, the bar is 100% of a bar width, and the gap plus the bar are 250%, so half the bar is 50%/250% or 0.2.

You can add duplicate Max and Min series, convert them to XY series, and format them with no marker and error bars of ± 0.2, to get matching end caps for the whiskers (high-low lines).

Excel 2007 (and Later?)

The protocol in Excel 2007 is similar. Create the line chart.

Add up-down bars and high-low lines using the Chart Tools > Layout tab of the ribbon. Don’t panic if the Mean and Median markers disappear. In Excel 2003 markers always appear in front of the up-down bars, and lines behind them, but in 2007 if the series has lines, then the markers stay behind the bars with the lines.

Remove the series lines, and the markers appear.

Change the markers, and use the dash for Median. The dash gets wider in 2007 than in 2003, so it looks silly sooner.

As in 2003, the median series can be changed to an XY type, moved to the primary axis, and given horizontal error bars. Adding error bars in Excel 2007 is a trick, though. You have to add them using the Chart Tools > layout ribbon tab, and this adds both vertical and horizontal error bars, with the Format Vertical Error Bars dialog showing. Clear the dialog, select the vertical error bars, and press Delete. Then select the horizontal error bars, which have a default length of 1, press Ctrl+1 to open the Format Horizontal Error Bars dialog, change the value to 0.2, and remove the end caps.

Likewise, you can add duplicate Min and Max series with error bars, to provide end caps to the whiskers (high-low lines).

I pointed out the generally poor ability to judge areas, and proposed a bar chart alternative that provides not only a qualitative3 display, but also a quantitative display.

The user posted the following sample data showing the hourly costs of operation of three conveyors, broken down into four constituent cost elements.

The user envisioned a display of pie charts similar to the following. Conveyor 2 has the greatest costs. Conveyor 1 has the lowest costs, only 35% that of Conveyor 2, and Conveyor 3 has nearly the same costs, 43% those of Conveyor 1. If we make the area proportional to cost, the diameter of the pies will be proportional to the square root of cost, so Conveyors 1 and 3 have pies with diameters 59% and 65% as large as that of Conveyor 2.

You can see that Conveyor 2 is the most expensive to operate, and you may get a sense that conveyor 1 is the cheapest. But it is at best a qualitative view. You cannot easily compare the magnitudes of the operating costs of the conveyors. You can roughly judge that Power is a somewhat larger proportion of Conveyor 2’s operating costs, but you cannot compare the relative costs of operator and parts.

The ineffective comparisons offered by the pie charts can be fixed by using a single bar chart. In fact, a simple clustered bar chart is sufficient, with one cluster per conveyor showing the total and constituent costs of operation. We can readily compare the total costs of operation, and we can see that the costs of operator and parts make up only a tiny and nearly invariant portion of the total.

Some labeling of the bars can help if we need to show precise values or percentages. The total bar labels below show simply the corresponding values, while the percentages of the other bars have been computed in the worksheet, and added to the bars using Rob Bovey’s free and indispensable Chart Labeler add-in.

Below are two radar plots of site statistics, Site Visits on the left and Page Views on the right, orbiting around the origin according to time of day. Midnight is at the top of each chart, noon at the bottom, and the curves proceed, well, clockwise around the charts.

What can we say about the information in these two charts? Well, Saturday and Sunday are close to the center, so they must have less traffic. Also, for the weekdays, the curves bulge at the bottom, so daytime obviously gets more traffic than nighttime. The times in the SiteMeter data corresponds roughly to my own time of day here on the East Coast of the US, since I get around 50% of my traffic from the US, though I don’t offhand know the East Coast vs. West Coast split.

The radar charts certainly look useful at first glance. They are multicolored, which is good, and there is a kind of satisfaction of feeling that the cyclical nature of time (Sunrise, Sunset) is reflected by the cyclical nature of the displays. Because the values are represented by radial distances in the radar charts, it is difficult to compare different times, and it is not easy to follow hour-to-hour trends in the data.

Another problem is that the larger values are at the bottom of the chart, which is counterintuitive. This could be corrected by putting midnight at the bottom of the chart.

This was easier than it seems: I’m using a combination chart, with a hidden radar series providing the spokes, and the daily traffic data plotted on hidden X and Y axes, using trig functions. All I had to do was reverse the directions of the hidden X and Y secondary axes, and change the category labels of the hidden radar series.

Maybe this helps, maybe it doesn’t. The larger values are higher in the chart, but the variation in the orientation of radial distances is not alleviated. We’re also conditioned to see midnight on the top of our clocks (noon too, on a 12-hour analog clock). The radar chart is the source of cognitive dissonance.

The inferiority of the radar chart becomes even more evident when compared to line charts showing the same Site Visit and Page View data (below). The charts start at midnight on the left and end at midnight on the right, and time flows smoothly from left to right. Instantly more detail is apparent than in the radar charts, and further details reveal themselves on closer viewing.

While the radar charts  show a nondescript jellyfish-like bulge during the day, the line charts show a two-humped Bactrian camel appearance. There is a big hump just before lunchtime on the East Coast, and a smaller hump in mid- to late afternoon. The first bump can be explained by people starting to zone out before lunch, and surfing the web. The afternoon bump is a bit late for the West Coast pre-lunch traffic, but the West Coast pre-lunch crowd may be partially obscured by folks on the East Coast killing a few moments before leaving work for the day.

A smaller bump is evident in the early morning in the Site Visits chart (and you can imagine a similar bump hiding in the Page Views chart). Could this be the pre-lunchtime bump from Europe?

The Friday curve is interesting. It begins the same as the other weekdays, but it lags by midmorning, and by evening it drops to the Saturday level. Obviously people are trying to finish up their week’s work, and aren’t surfing as much. They may be leaving early, and nobody is browsing technical web sites on the weekends.

By midafternoon Sunday, traffic starts to pick up. I suspect this is due to my Asian visitors accessing the site. This traffic is usually drowned out by Europe and especially the Americas, but on Sunday evening (Monday morning to midday in Asia)  it’s the major source of visitors.

That’s been my most popular blog post, and probably the most commented. Many of the comments ask about modifications to the technique to achieve different effects, and mostly these are achieved by carefully adjusting the arrangement of the data. One such request was, How do I add a target line that reaches to the edges of the chart?

The answer isn’t too complicated. Set up the data as shown below.

Using the technique in Clustered-Stacked Column Charts, create a chart using the first five rows of data.

Copy the last row of data (including the label in the first column), select the chart, and use Paste Special to add the data as a new series. A new column series will appear stacked on the rest.

Right click the new series, choose Chart Type, and select the line without markers option. Excel helpfully adjusts the secondary category axis, making the labels visible and making the value axis cross between categories.

No matter, we fixed this once before during the protocol, we can fix it again. Format the secondary category axis so it again shows no category labels, and so the axis crosses at the categories, not between them.

The data I used is shown below. A stock chart needs a column of dates plus four columns of price data. The price data has to begin with Open and end with Close. High and Low can be in any order in between, but conventionally High is listed first. I have included two index columns, one within the same range as the stock prices, one in a much higher range.

I have an extra row above and below the data. I prefer to have my vertical axis cross on a date, not between dates. This is fine, but the first and last candlestick in the chart are truncated (even if there are more dates on either side) unless you explicitly include an extra date before and after the data. Also, I’ve hidden rows 12 through 35 so the data doesn’t take make you scroll down two screens.

Excel 2003 Stock Charts

Select the first five columns and create a stock chart. I’ve applied fill colors to and removed the borders from the up-down bars (I always forget whether the white and black bars mean up and down, or vice versa).

If you want you can change the width of the up-down bars. Format any of the four series: on the Options tab of the Format Series dialog, you can change the Gap Width. Note also that you can add or remove the high-low lines and up-down bars.

Add the index data to the chart. The easiest way is to select the chart area or plot area of the chart, find the colored highlights indicating the chart’s source data range (see below), and resizing these ranges with the mouse. You could also copy the added data and paste in into the chart (I prefer paste special), or use the Source Data dialog to define a new series.

When you widen the blue outline to the right, the green outline also widens to the right. I stretched it by one column, which added the Index1 data to the chart. We can’t see the series itself, but we do see a profound distortion of the up-down bars.

Below I have selected the Index1 series and changed its chart type to Line. As you will recall from reading Stock Charts and Other Line Chart Tricks, a candlestick chart is just a dressed up line chart, with the series lines hidden, and high-low lines and up-down bars added to the chart. The Format Series dialog above is a reminder of this fact.

The up-down bars are messed up because the Index1 series is just another line series in the chart. Since it’s the last one in the plot order, it becomes the close end of the up-down bars.

Fix the Up-Down Bars Using the Secondary Axis

There are two ways to fix this mess. The first way is to exile the Index1 line series to the secondary axis.

If we were plotting Index2, with its values so much greater than those of the stock prices, we would fine tune the secondary axis scale and call it a day.

However, Index1 falls within the same range as the stock prices, and it would be best to use the same axis scale. You can laboriously format the two axes to align their scales, then whenever the data changed, laboriously reformat the scales.

Or you can remove the secondary axis altogether, which forces the secondary axis series to use the primary axis. Remove the axis by selecting and deleting it, or by going to the Chart menu > Chart Options > Axes tab, and unchecking it.

The up-down bars and high-low lines apply to all series within a chart group. A chart group is the group of all series of the same type (i.e., line) and on the same axis (primary). Since the Index1 series in in the secondary chart group, it no longer affects the primary chart group’s features.

This technique is pretty clean, except it ties up the secondary axis, preventing its use for plotting of other data (e.g., Volume) with different value ranges.

Fix the Up-Down Bars Using an XY Type Series

The second way to fix the messy up-down bars created by adding the Index1 series is to select the series and change its chart type to XY. Excel automatically assigns the XY series to the secondary axis, and we are given two misaligned axes.

If you reassign the XY series to the primary axis, it uses the existing axes, even though the axes were defined by the stock chart (line chart) series.

Using an XY chart has the benefit of putting all of this data onto the primary axis. This is useful if there is more data, such as Volume, that you would want to plot on the secondary axis.

Another benefit to using an XY series is that you can plot data with different dates than used by the stock (line chart) data. All line type series in the chart group have to use the same X values.

Excel 2007 Stock Charts

The Excel 2007 stock chart is constructed much like the Excel 2003 version.

Adding the Index1 series is accomplished the same way, by stretching the highlighting around the chart’s source data range. You can also copy and paste the data, or use the Select Data dialog, but I find this dialog much more confusing than the Excel 2003 version. However, the result shows the same distorted up-down bars as we saw in 2003.

The Index1 series can be changed to a line chart type series just like in 2003.

Fix the Up-Down Bars Using the Secondary Axis

Index1 can be assigned to the secondary axis, just like in 2003.

The secondary axis can be removed, just like in 2003.

As before, the first technique is good if you don’t need to use the secondary axis for anything else.

Fix the Up-Down Bars Using an XY Type Series

When we try the second technique, changing the type of the Index1 series to XY, we have a rude surprise. Apparently stock charts in Excel 2007 don’t realize they are just gussied up line charts. Whereas we could combine XY and Stock types in a single chart in Excel 2003, Excel 2007 disallows this combination.

Once you get over your initial shock, remember that a line chart can be dressed up with up-down bars and high-low lines to become a stock chart, much like Cinderella can be dressed up to become a princess.

Convert the chart from stock chart to line chart.

The steps to add up-down bars and high-low lines have been changed drastically in Excel 2007. These features are no longer conveniently formatted in the Format Series dialog. Instead you need to navigate through the contextual Chart Tools tabs of the ribbon. You can find the controls off to the right end of the Layout tab.

The gap width control for the up-down bars has been moved as well. In fact, as far as I can tell it has been completely removed from the user interface. But in a moment I’ll show you how to change this with a little bit of VBA.

Here’s our line chart with the stock chart features:

Hide the series lines on all but the Index1 series, and it looks familiar.

Change the chart type of Index1 to XY.

Reassign Index1 to the primary axis.

There is our combination XY-Stock chart. Using the XY series allows plotting of data with different dates than the stock prices, and leaves free the secondary axis for other data.

Changing the Gap Width of Up-Down Bars in Excel 2007

To change the gap width of the up-down bars, you need to apply a little VBA. But I’m going to try to make it less scary. First, select the up or down bars, then click on the Chart Tools > Layout or Format tab, and look at the Current Selection box in the top left of the ribbon.

It says Up-Bars 1, and the 1 refers to Chart Group 1. Open the Visual Basic Editor (VBE) by pressing Alt+F11, open the Immediate Window by pressing Ctrl+G, type the following line, and press Enter. Note that ChartGroups(1) refers to the chart group index we found above. Experiment with the gap width value until the bars look good. For this particular chart I didn’t have to change the gap width, but I’ve had others which required it.

The problem with Funnel and Pyramid charts is that they usually suffer from a misapplication of the funnel metaphor, and Chandoo’s is no different. In this incorrect metaphor, a funnel is used to show an ever narrowing amount of something (sales, in Chandoo’s example) the further along the stream the process goes. The quantity decreases because some of it is removed from the stream or blocked from proceeding. In a real funnel, there is no loss of this quantity, in fact the flux of the quantity past a point must increase to account for the decrased opening size. But nothing is lost.

In fact, a better representation would be a Sankey Diagram (see below), which accounts for the losses through the process. But that’s a discussion for another day.

One can still use a bar chart to represent the quantities at each phase of a process. One improvement is to move the bars to one side of the chart. This gives better resolution to the values (wider difference on the chart for the same difference of the measured quantity). This also makes it easier to compare relative values. Because of the shared baseline along the left edge of the chart, it’s easier to judge that in Phase 2 and in Phase 3 the values are respectively a bit over and a bit under 50% of the value in Phase 1. In the tornado chart above, one could not be so confident in these estimations.

In Tornado Charts and Dot Plots I described a different type of chart, variously called a tornado chart or a funnel chart, which compares two populations by plotting bars either to the left or to the right of the middle of the chart. The symmetry is appealing, but the ability to compare two bars that strike out in opposite directions is compromised. I suggested using dot plots instead. Here is a rudimentary dot plot of Chandoo’s data:

If we are satisfied with the resolution of the original funnel chart, we can still benefit from the common baseline by using left-aligned bars half as wide as those of the funnel chart. This also saves a lot of space in a crowded report.

One could also make a column chart of the data, but the category labels either must be rotated or shortened, reducing legibility in both cases.

At the outset I said that Chandoo’s protocol was “okay”, but he could have used a different bar chart approach to make it better. Basically, a funnel chart is made of two series, one showing the values (orange below) and the other normally hidden (green) which helps to position the value bars.

When I make charts like this, I make floating bar charts, which are stacked bar charts, with a dummy series that pushes the value series into position. Below I’ve offset my green and orange bars to illustrate the construction. I can hide my dummy series by formatting with no borders and no fill.

Chandoo chose instead to use clustered bars, as shown below, with the orange bars including the data plus the (green) offset into a total value. The green bars are then moved in front of the orange bars and hidden by formatting them with a fill color that matches the plot area background.

As soon as somebody tries to read values off the chart by mousing over the bars, the superiority of the stacked bar approach is evident. Below I’ve moused over one of the value bars, and we see it’s from the Funnel series and the value is 21,180.

If I try to mouse over one of my dummy bars, the cursor can’t see it, because it has no fill, and instead all that is detected is the plot area.

In the clustered bar approach, the exaggerated value of the Total series is displayed (53,593), and if this point is selected, the covered end of the bar is outlined.

The hidden points can also be detected, as in the Dummy point below with a value of 32,408.

Sometimes it helps to use fancy formatting or designs to make a chart more noticeable. But in a business report, if the data is important, it does not need fancy effects to be noticed. Selecting a funnel chart, even using the improved stacked bar approach, is still a compromise which decreases the effectiveness of the chart.

No surprise that almost half is spent lighting the display.

But wait, that looks a lot closer to 50% than 43%. Let me see what I get using their numbers.

Looks the same. Let’s do a quick check of the numbers. 43% + 21% + . . . look at that, 90%! Often the numbers are off by 1% or thereabouts, due to rounding effects in the labels (but see Pie Chart Rounding in Excel for a discussion of rounding errors in Excel pie charts). But in this case they’ve obviously left something out.

All the well-intentioned guidelines for charting say to use a pie chart to show how parts make up the whole. And yet, here we have only some of the parts making up 90% of the whole. The basic premise of a pie chart is violated by omitting this 10% of the total energy usage.

Pie charts are supposedly the best way to show proportions of this sort, although they are not really so good even at that. People mistake familiarity with effective information display.

The same data is shown as effectively in a bar chart, either with the original values:

or with the missing 10% accounted for:

These charts are not distorted by exclusion of some of the data; no basic assumptions of bar charts have been violated.

In addition, at 320×221 pixels, the original bar chart uses only 54% of the space as my original pie chart (426×306 pixels).

The numerical data is Minimum Inhibitory Concentration (MIC), the smallest amount of an antibiotic required to prevent in vitro bacteria growth. Presumably it’s in units of mcg/ml, but I could not verify that. Gram staining indicates whether the bacteria picks up a dark blue or violet color (positive) or a faint pink (negative) when exposed to Gram stain. This is a quick-and-dirty way to separate bacteria into two classes.

The first step in any analysis is to plot the data. Here I’ve created a bar chart with MIC on a logarithmic scale.

The larger the concentration, the less effective an antibiotic is at inhibiting bacteria growth, so the longer bars above indicate lower effectiveness, which is counterintuitive. My next step then was to reverse the direction of the log scale, so that longer bars correspond to higher effectiveness.

That’s better, but it’s still not very easy to read. With so many categories (bacteria species), the bars are too narrow to be clearly resolved.

I decided to try a panel chart to separate the colors, with one panel for each antibiotic. Here is the first cut.

We see the alphabetical sorting issue in this chart. There are a number of ways we could sort the data, and probably the most arbitrary is by alphabetical order of the category labels. What if I sort by effectiveness of Penicillin? (Once the chart is created, the points are sorted by sorting the original data table by the desired column.)

That’s more meaningful. Here is the data sorted by effectiveness of Streptomycin:

Now sorted by Neomycin effectiveness:

I thought the panel charts could be simplified by using the Overlapped Bar Chart – Longer Bars in Back approach of Robert Kosara of EagerEyes. This shows the data ranked by penicillin effectiveness.

There are problems with this kind of chart. First, I had to choose an uglier green, because the original green was too hard to distinguish from the blue. Second, It’s difficult to get over the ingrained misunderstanding that the bars are not stacked, but shown shorter in front of longer. Also, there are several places, circled below, where two bars coincide (or nearly coincide), making it impossible to identify all three values for each microbe.

These attempts so far show the effectiveness of the antibiotics well enough, but they omit the Gram staining behavior.  I decided to shade the rows next to Gram positive bacteria, and use a dot plot instead of a bar chart to show the data.

There’s that alpha sort issue again. With this arbitrary sort, it’s very hard to see any correlations in the data. But when we sort by Penicillin effectiveness, we see much more information. When sorted in this way, the Gram-positive bacteria are seen as those against which Penicillin is most effective.

Sorting by effectiveness of Streptomycin and Neomycin (omitted to keep this post reasonably long) is good for comparisons among the antibiotics, but loses the impact of the correlation between Penicillin effectiveness and Gram staining.

This type of chart readily translates to a gray-scale version.

This is the chart I submitted to the contest. What could I have done better?

Flojoe’s original data is shown in bold text in the table below, and I’ve added some calculations in regular text. The row called “Base” refers to the prior year’s value. Cell B3 is zero, cell C3 links to B5 (=B5), and cell D3 links to C5 (=C5). The row called “Growth” refers to how much more the current year is over the prior year, so cell B4 contains the formula =B5-B3, and this is filled to column D.

The row called “Labels” will be used to label the “Growth” series once it’s plotted. Cell C7 contains the formula =TEXT(C6,"0%")&CHAR(10)&"Growth", which is copied into cell D7. The formula tells Excel to use the value in cell C6 formatted as a percentage with no decimals, plus a line feed, plus the word Growth.

To start, create a line chart using the green highlighted range, with series in rows. (The yellow range will be used later to label the chart.) Simple enough, but it doesn’t really show anything, yet.

Start to separate the data into panels by formatting the SOW Count series so it is plotted on the secondary axis.

Right click one of the other series (I’ve used Growth, but Base would work as well), choose the Chart Type command, and select a stacked column type.

Right click the other remaining series, and again convert it to a stacked column type. Growth is now stacked atop Base.

Now to format the series. I made the SOW Count marker into a square so it stands out more, I made Base light blue, and I made Growth green. I changed the gap width on the columns to 100 and removed their borders.

It’s time to make the chart into an actual panel chart by formatting the primary and secondary Y axis scales. Make the primary (left) Y axis go from -250,000 to +400,000, and the secondary (right) Y axis go from zero to 130. Each axis has five divisions below and eight divisions above the category axis, so the tick marks and gridlines are in perfect alignment. This sometimes takes a bit of trial and error; see Align X Axis to Y=0 on Two Y Axes (VBA) for a similar alignment.

Format the category (horizontal) axis so that the labels (years) are in the Low position (below the chart) and remove the axis tick marks.

Add titles to the primary and secondary Y axes, and apply custom number formats to hide the scale labels where they are not needed. For the primary (left) Y axis I used a custom number format of $#,##0;;$#,##0 which tells Excel to use $#,##0, a format with a leading dollar sign and commas for thousands separators, for positive numbers and zeros, and ignore negative numbers and text. For the secondary (right) Y axis I used a custom number format of [<=50]0;;; which tells Excel to use 0 for numbers less than or equal to 50, and to ignore anything else. For more details, see Number Formats in Excel.

Finally the Growth series needs labels. I use Rob Bovey’s Chart Labeler utility, which supplies a missing capability in Excel: the utility allows you to enter labels into a range of cells, then apply these custom labels to a chart series. When you install this utility, it adds a Chart Labels command to the Excel Tools menu. Choose the Add Labels submenu command, select the appropriate options, and the selected series gets its labels.

The near-finished panel chart is shown below. The SOW Count and SOW Revenues are shown in separate panels so the behavior of each can be observed without one series and its axis scale interfering with the others. Not only revenues, but also year-to-year growth of revenues, are clearly displayed.

The horizontal gridlines serve the same purpose as the Y axis tickmarks, so the tickmarks can be removed.