15  Graphing Tips

This chapter includes a few short sections about “advanced” CODAP graphing tips.

15.1 Re-ordering categoricals

If you make a graph with a categorical attribute, by default CODAP orders the alphabetically. Sometimes, this is maddeningly unhelpful!

In many cases, the solution is quick and easy: grab the categorical value by the name, and drag it where you want it.

In this live example, we’re looking at age (and mean age), grouped by marital status.

The order doesn’t make sense. So re-order the marital-statuses by dragging the labels (like Never married) up and down.

15.2 Fusion Power: Bar charts

If you have made a categorical graph, you can change the default “dot” display into a bar chart, which may be more familiar.

Also, sometimes, when you have created summary values (such as the means of something for several groups), it may be effective and appropriate to display those aggregate values as bars.

Your basic bars

Make bars using the “configuration” palette and choose the option Fuse Dots into Bars.

For example, using the same data with marital status (but not age), here is a “before” picture:

…and the “after” picture.

Making proportion bars

As you can see, you can then scale the chart in various ways.

The “percent” scaling is particularly useful. If you plop a legend attribute into the graph, the bars will split, like this:

It looks like there are a lot more divorced women than men in this sample. But what is the proportion? Choose Percent in the configuration palette:

Aha! Now we can see the proportions of men and women for each marital category.

Tip

Once the data are in bars, you can select an entire set of cases by clicking on its bar.

Controlling bar length with a formula

One last feature. What does the Formula option do?

Using the same dataset, we have Marital_status on the vertical axis and click Formula in the configure palette. That invokes the formula editor.

We enter mean(Age) and see bars whose lengths are the mean ages of the people in that category. And it makes sense! Widowed people are older, never-marrieds are younger.

Your turn

Use the live example below to practice any of these techniques:

15.3 Showing data directly as bars

CODAP generally shows cases as dots. But if you are plotting one attribute, the configuration palette will let you plot numeric data as bars.

For example, if we get some NHANES data and put Height on the vertical axis, and choose Bar for Each Point, we see this:

Which is a bit confusing but kind of cool; each bar is the height of an individual. But you can sort the data by Height in the table, and miraculously, the bars re-order as well:

Two caveats (as of January 2025)
  1. Remember that you can only do this if there is only one numeric attribute on the graph. If you already have two, the Bar for Each Point box doesn’t appear.
  2. This sort-to-reorder-the-bars feature sometimes doesn’t work. You have been warned!

Bars often work great with summaries!

That last, sorted graph works well in that it gives us a different kinda-cumulative view of the distribution of heights. But it’s an exception; mostly, dots work better to represent individual cases.

But part of what we do to alleviate being awash in data is to group data and summarize, removing detail so we can see an overall pattern.

In an early lesson, we summarized heights by age and plotted one do—one mean height—for each Age (5 to 19) to see how we grow. If you wanted to see that with bars instead of dots, use the live example below and do the following:

  1. Make a new graph
  2. Put meanHeight on the vertical axis.
  3. Choose Bar for Each Point from the “Configure” palette. This creates a messy graph!
  4. Click on Age and choose Sort Ascending. Ta-daa!

Because we’re looking at Height, and bars, by their nature, extend to zero, it’s as if we’re looing at a row of (rectangular) people. You can really see the difference of heights, in proporrtion, between five- and nineteen-year-olds.

Note, though: this still only works of there is one attribute on the graph. If you put Age on the horizontal axis, you get dots again.

15.4 Histograms

How do you make a histogram in CODAP? You might expect (correctly) that you will fuse dots into bars at some point, but first you need to bin the data.

Note

I am making a distinction here between bar charts and histograms. We just made bar charts when we fused dots with categorical values. Notice that they had spaces between the bars.

In a histogram, the data are numerical, and there is no gap between the bars.

The live illustration below has 1000 heights of people between 25 and 35 years old. We want to make the dot plot into a histogram. Do the following:

  1. Select the graph and open the configuration palette
  2. Choose Group into Bins.

The graph adjusts into bins; notice that each bin is 10 cm wide.

  1. Again in the configuration palette, choose Fuse Dots into Bars.

Now you have a histogram! Use the configuration palette to experiment with the other controls, such as the bin width.

You would think that if you put Gender on the vertical axis, that it would split the histogram. But as of February 2025, you will only get the binned-dots version—not fused. Still useful, though!

15.5 Connecting the dots

When we teach graphing real data by hand in secondary school, we often have to work to get students to stop connecting their dots with lines.

But sometimes, it’s appropriate. A time series is such a situation, or any time the thing you’re plotting might possibly be changing continuously. Also, sometimes you might need lines to emphasize the direction of a change.

To connect your points, use the ruler palette and check the Connecting Lines item.

To practice, we’ll look at time-series data from a site in the Blodgett Forest in 2017.

In the live example below, you see the temperature 10 feet above the ground and also the temperature 140 feet above the ground, every 15 minutes for three days at the end of September.1

You can see the overall pattern: it gets warmer in the day and cooler at night. What else you can see?

Then, change the plot:

  • Click the graph to make the palettes appear.
  • Go to the ruler palette and choose Connecting Lines. You’ll see the lines appear and the points get smaller.
  • You can make them larger or smaller still—or even make them disappear!—using the paintbrush palette.

It’s OK to connect the dots because the temperature would still have a value in between the points, and a linear connection is a reasonable guess.

Extra task to try:

  • Plot T_soil against T_10ft.
  • Plop stringDate into the graph and expand the dot size (so you can see the color).
  • Look at the pattern and try to understand it! Especially: why is it tilted that way?

Notice that it’s still appropiate—in fact, almost essential—to connect the dots to see the pattern.

The big connecting-line gotcha

The next dataset shows 18 months of high temperatures from Chicago and San Francisco.

We’ve already colored the points by station. All you have to do it connect the points.

Dang! What’s that additional line? Ah: it’s where the last point of Chicago connects to the first point in San Francisco.

The fix is easy: drag station to the left, that is, group by station. The lines connect points in the same group.

  1. The decimalDate is the day of the year, with “.0” being midnight local time (PDT). The when attribute shows time and date—at Greenwich, which is 7 hours ahead of Blodgett in California.)↩︎