Tags:

# What is the point of a rose chart?October 21, 2013 2:44 PM   Subscribe

When is it more appropriate to use a rose (or radar) chart vs. a bar/line/pie chart? Bonus points if you can point to studies showing why some types of charts are better than others for understanding information.

On this page (near the bottom) there is an example of a rose chart. Here is a more complicated one - possibly the first - created by Florence Nightingale.

I understand how to read these, but I don't understand the proper place to use them. I understand that generally pie charts (or donut charts which don't even get me started) are used to compare a thing in relation to the whole. I understand that bar charts are good for comparing things to other things, and line charts are used to show change, usually over time. I get why you'd show change over time via the y-axis and change in quantity via the x-axis.

But if the point of a chart is to display information for people to parse quickly and easily, then I don't see the point of a rose chart other than aesthetics.

Can someone explain this to me?
posted by nushustu to education (11 answers total) 19 users marked this as a favorite

There is a great thread on this type of chart (and the Nightingale chart in particular) at Edward Tufte's website. Tufte's comment links to this study that also goes into some detail.

Tufte's books cover visual data in gorgeous detail and I highly recommend them to anyone and everyone, if for no other reason than their beauty.
posted by jquinby at 2:53 PM on October 21 [3 favorites]

I like them when there's something periodic that you're displaying, like average temperature by month; December to January is just as long as January to February, but usually they're on opposite sides of a bar or line graph.

In my field (neuroscience), we use them to display information about phase, which is pretty common. (Waveforms are roughly sinusoidal, so each point on the wave has a corresponding degree/radian: peak at 90 degrees or pi/2 radians, trough at 270 or 3pi/2, etc.) If something interesting happens at, say, the peak of each waveform, it's easiest to see on a rose plot, something like this. (I have no idea what it's measuring, but it's saying that something interesting is going on at 0 degrees or so.)

This makes it much easier to see patterns that divide the total rotation in half. So if something interesting happened every six months, like alien abductions, it's easier to see it when the relevant bars are directly opposite each other rather than at two seemingly-unrelated points along the X-axis of a line plot.
posted by supercres at 2:54 PM on October 21 [4 favorites]

Here's a much better example, with link to paper intact.
posted by supercres at 3:00 PM on October 21

If the rose chart can organize the data in a way that is congruent with the categorization of the data — such as periodic time, phase or directional categories — then it may be easier or faster to draw the correct conclusion from the data at a glance.
posted by Blazecock Pileon at 3:06 PM on October 21

I think supercres and Blazecock have the right idea - they're for cyclical periodic things or directional things. So for example, if you want to show how value Y has changed over time, you'd probably choose a line graph. But if you wanted to show how the changes in Y tend to align with certain times of the year, you'd average all the Januaries, then all the Februaries, etc. and put that on a radar or rose chart.

The charts in the link you gave aren't very good examples of when to use that chart type. Why is Missy next to Rick and not next to Jack? Those names are just randomly scattered around the perimeter. There's not any extra periodic relationship between them that could be illustrated, so there's no point in doing a fancy radar or rose chart.

This may be sacrilege, but I don't actually like Nightingale's charts. I respect her pioneering work, and I know they convinced people to start taking battlefield diseases seriously, but I honestly think the data would have been better served by a bar or line chart. The dotted connecting line to link two different years is especially awkward.

Also bear in mind when choosing charts that most people choose a chart type based on aesthetics and don't really think about the best way to display the data, so don't just look and see what other people are doing. If you're stopping to consider this at all you're already ahead of the crowd .
posted by echo target at 3:34 PM on October 21

I don't really know how to draw a broad rule of thumb from the following, so I'll just be specific: modified rose charts are great for showing the difficulty of songs in the videogame Dance Dance Revolution (a full-body arcade game where the player has to step/jump on 4 footpads in time to a techno song). The different types of difficulty- speed, rhythm-complexity, intensity, predictability, acrobatics, etc, can each be represented by a segment of the circle.

The rose chart is great here, because the overall difficulty of the song can be read at a glance when scrolling through many songs. The harder a song is, the bigger the splotch in the difficulty chart will be, and depending on the skew of the splotch, you can see at a glance how difficult it will be in a particular skill area.
posted by pseudostrabismus at 3:35 PM on October 21 [1 favorite]

They are used in the oil industry when displaying trace data with azimuth.
posted by Houstonian at 3:44 PM on October 21

I have seen radar charts used to plot several characteristics (eg speed, quality, price), where they apparently are being used to give an eyeball estimate of overall desirability based on the shaded area. In this respect they're essentially bar charts wrapped in a circle. I don't know if the radar format makes that information more digestible or not; I suppose it reinforces the fact that there's no sequence involved, whereas a bar chart might suggest that there is.

But using a radar chart to plot periodic data makes a lot of sense.
posted by adamrice at 4:03 PM on October 21

(I don't think the Nightingale chart is an example of a rose chart, so maybe that is part if the confusion?)
posted by Houstonian at 4:13 PM on October 21

Wind roses are self-explanatory, right?
posted by scruss at 7:55 PM on October 21

I think they'll be extra helpful when you are comparing cyclic data for several entities, so that you'd have several "roses" next to each other. Then the extra cognitive effort required to read the first one is compensated by the ease of comparing one thing to two things.

The wind rose, for example. You could take wind patterns for one weather period (morning, for example, or maybe the entire year 2003) and easily see how they differed from another (evening, or every year for a decade).

This isn't the world's best example, but at least in about the middle of this page you can see two related charts together, and you can experience how easy it is to compare them.
posted by amtho at 11:45 PM on October 21

« Older I've taken cooking classes at ...   |  What should Australian propert... Newer »

You are not logged in, either login or create an account to post comments