I have seen routines to adjust the positions of individual labels until there is no overlap, but these have been rudimentary, and I cannot locate the specific one I’m thinking of on the author’s web site.

You can position the data labels manually: click once to select the series of labels, and a second time to select the individual label, then drag it into position. It is not easy to control the position closely using the mouse. Rob Bovey’s Chart Labeler has a feature that gives you closer control over positioning of data labels.

For the xy chart, I think it would be very nice to avoid the labels overlapping. I copied some vba code for attaching labels to the xy chart, but came with the same problem about overlapping. I was wondering whether you know how to solve it, and would be very grateful if you could share it.

If you and your lawyer friend Ken don’t mind, perhaps I could write about that macro some day. It’s not a complex topic, well, not for someone who has been around Excel and VBA as long as I have, but it’s a good way to discuss more uses of Excel and VBA together.

Nixnut -

I was suspecting that Naomi was using a select-the-point-and-paste approach, but I was thinking about something that used a bar chart and an XY point with a custom marker where the arrowhead goes. This keeps the shape of each arrowhead consistent despite the changing lengths of the arrows.

I’ve written VBA procedures that create a chart, plot X and Y based on two fields in a data set, then adjust a shape according to one or more additional fields, and apply the shape to the chart. I’ve done this to illustrate a vector field, that is, an array of arrows which have varying length and orientation. I’ve used it to make custom bubble charts, where the bubble is something too complex to render in a simple bubble chart (like a pie chart, but don’t tell anyone).

I did this the first time I used arrows and then Ken Klein, a brilliant lawyer who knows more about Excel than most Excel gurus, generalized what I did and wrote a macro for me to use in the future. Anyone who wants a copy of the macro can send an email to me at naomi at nbr-graphs.com.

Thanks for chiming in. I feel the oxygen level increasing.

The arrow chart is effective, because the colors and arrowheads both indicate the direction of spending change. This is essentially all the information that I had in my line chart, but the arrow chart eliminates a couple of drawbacks to the line chart. As you and others have pointed out, the line chart as drawn implies linear growth over the ten years, and the way I’ve executed my line chart, the labels add a lot of clutter.

I’m already thinking of ways to have Excel draw bars with arrows…

The percent change can be added as labels without needing the segment part of the label in the line chart or we could add a similar chart based on percents if desired. Since this emphasizes the beginning and end points and the amount of change, it’s not assuming linear growth.

The main advantage would be seen if we had many more than four categories. In that case the lines might all cross one another and it would be hard to determine which category a line was associated with, but the arrow chart allows any number of categories. On the other hand, lines can accommodate more than two time points which this chart cannot.

Both Jon’s line chart and Tony’s table are cluttered with unnecessary percent signs, dollar signs, or B for billion. I’d prefer an axis label saying billion dollars with the labels just being 0, 10, etc. The numbers are clearer when not surrounded by the clutter of $ and B. In the table I’d keep the % sign in the column label and with 100% but eliminate the others.

The second chart I think is right on. It seems to me that the whole point of the original chart is to evaluate if there is correlation between poverty and graduation rates. An x-y chart is the proper tool to show that. The specific state information, while perhaps interesting, is secondary to the issue of correlation and doesn’t need to be shown.

I also am always analyzing the data, looking at many more views than ever make it into any deliverable form. In some of my posts on this blog, I try to show some of the paht through the data that ultimately results in a display.

The line chart is pretty good, but a bit cluttered with the labels. It does show the segment changes much better than the pie does. Even though the table takes up less space, the line chart is able to show the trends much better. Maybe a combination of chart and table (beneath the chart) would be optimal?

In the second section, I had noted that the print version of BusinessWeek showed the bars in one column versus the three columns online. As I had suggested, the only change I would make to your bar chart is to sort descending versus alpha for the poverty rate when the graduation rate is the same.

The heat map can be easily done using Many Eyes if you have the data.

The XY charts are an excellent addition and tell more of the story. I am always analyzing versus presenting data. I want to know the story, trends, correlations, etc. and not just a percent change. Also, the XY chart using the states is a nice variation. What are NJ, VT and NH doing so right compared to NE, which is atrocious. A 45% high school graduation rate, really???