Jon’s Radio Comments

October 30, 2006

Scaling the Tufte effect

Filed under: Uncategorized — jonsradiocomments @ 12:32 pm

The original item is here.



  1. Tufte’s graphic is truly awful. Pretty, yes, and efficient — but deceptive. To look at it, one would expect that the graphs are on a common axis, and to scale. They are not: the lines cross on a common scale.

    On the web, his graph could be fixed by the mechanism you show: by letting the lines cross, but by picking the delta relative to which the sorting should be performed.

    Comment by John Merrill — October 30, 2006 @ 2:39 pm | Reply

  2. Well, John…note that in context, Tufte’s sparkline graph makes more sense (again, IMO). As he notes in the forum, “For most presentations, this table [note: a different, text-only one] with its structure and reporting of standard errors will be the best way to see the cancer data. The table-graphic below [note: the sparkline graph], however, gives an idea of survival time gradients for each cancer”

    So the sparkline graph is designed to show off the gradients within a given row, rather than comparison across types. Different presentations enhance different views of the data, and it’s always a give and take.

    Comment by Ken Kennedy — October 30, 2006 @ 3:10 pm | Reply

  3. > Different presentations enhance different views of the data

    Yes. And the real point I’m making here isn’t about the pros and cons of this particular visualization technique, but about how we become aware of, and able to make effective use of, a broad spectrum of visualization techniques appropriate to a range of communication needs.

    Comment by Jon Udell — October 30, 2006 @ 8:38 pm | Reply

  4. For an interesting discussion of the future of bringing info viz to the masses, check out Fernanda Viegas talk on “Democratizing Visualization” at the recent IDEA2006 conference:
    (Her talk was on day 2)

    Comment by Karl Nelson — October 30, 2006 @ 8:52 pm | Reply

  5. FWIW, This item was picked up by both digg and reddit. What interests most me is the comments on reddit:

    None are about Tufte, or Python, or data visualization techniques. All (so far) are reactions to Tufte’s interpretation of the data, and how that relates to (I’m presuming) people’s own experiences, or the experiences of their friends and loved ones.

    When data visualization works properly, it becomes invisible — a non-issue. Attention focuses on the /story/ that the data are telling.

    Comment by Jon Udell — October 31, 2006 @ 1:33 am | Reply

  6. Well, actually, the reddit comment about pancreatic cancer is a great example of why I despise this particular graphic. Yes, only three per cent of the those diagnosed live ten years — but only four per cent live five years. The means the marginal death rate drops sharply. Tufte’s graph hides that. In addition, it’s clear from the comments that people don’t realize that the lines cross, and that many of the numbers in the last column are out of order.

    Sometimes, a graphic makes data into fiction. Surely that’s worse than nothing?

    Comment by John Merrill — October 31, 2006 @ 4:18 am | Reply

  7. > …the marginal death rate drops sharply. Tufte’s graph hides that

    So here’s the deal. Show us your interpretation of the data. And then give us the tools to do it for ourselves.

    I’m not arguing for or against the graphic. I’m arguing for a culture in which we can riff on data visualization, and trade those riffs, and arrive somewhere new.

    Comment by Jon Udell — October 31, 2006 @ 5:20 am | Reply

  8. I wonder whether richer frameworks like Adobe FLEX might be able to extend their graphing components with sets of sparkline graphic generators. It seems to me that your approach could certainy be knocked up with some of the current components (tabs for your different view filters, a line graph for each of the rows and a repeater or somesuch to generate multiple rows). You can always add an extra tab to view the raw data in a more standard data table format as well.

    Comment by Mark Thristan — October 31, 2006 @ 2:26 pm | Reply

  9. I think the biggest problem with most visualisations on the web today is not their lack of interactivity, but the lack of easy access to the data. As long as you can get to the data, you can always use your favourite tool.

    (I was also pleasantly surprised to see that you’re using a wordpress template that I designed a while back)

    Comment by Hadley Wickham — October 31, 2006 @ 7:28 pm | Reply

  10. Because the data-presentation (and analysis) techniques here focus tightly on “visualization,” would this be an opportune context for small screencasts to be made, in which the data presentation could include a “movie” which uses the data analyst’s voice and “laser pointer” (ie, cursor) to elaborate on the more difficult or ambiguous portions of the visual presentation? In addition, the movie could include zoomed (in or out) portions, in order to tell a more complete story.

    Similarly, readers of the data could make small screencasts which highlight their questions/problems with the interpretations of the author(s).

    One purpose of visualizing data is to disambiguate the interpretations as much as possible… when this doesn’t succeed, perhaps another visual technique, screencasts, would be useful (along with access to the original data, of course, to allow alternative analytical techniques).

    Comment by Mike Lougee — October 31, 2006 @ 11:11 pm | Reply

  11. A small nit off the main branch, but I don’t see how survival rates can improve over time, as in the case of thyroid cancer in this chart. Perhaps I am missing something here.

    Comment by Andrew Binstock — November 8, 2006 @ 7:46 pm | Reply

  12. I think that has to be due to the margin of error.

    Comment by Raj — November 9, 2006 @ 7:47 pm | Reply

  13. Flipping the data around creates a graph where the different cancers and their survival rates are easily compared: Survival rates over the years are easy to see/compare as well.
    This data flipping also works for the different types of cancer:
    Thanks to Jon for making the data available via a separate URL.

    Comment by Ken Chomic — November 27, 2006 @ 8:24 pm | Reply

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Create a free website or blog at

%d bloggers like this: