How to graph this data?

June 13, 2009 3:33 PM

I'm looking for a way to graphically represent data that is basically "yes" or "no".

I am writing a research report on case study of a specific pathology where, through a series of 8 treatments, I have measured results of "special tests". There are 3 tests that were measured each treatment, the results of which are either positive or negative.

I'm looking for a way to graphically represent these data points to include in my research paper, as the test results show the progress of the treatments (a negative test implies resolution). Any thoughts?

Here's my data, if it helps... (not sure how it will render):

Session Phalen's test Prayer test Tinels sign
1 1 1 1
2 1 1 1
3 0 0 1
4 0 1 1
5 0 1 1
6 0 0 1
7 0 0 0
8 0 0 0

0=negative result, 1=positive

posted by nitor to computers & internet (7 comments total)

I've seen this done two ways:

If you're going to draw the grid lines, you can simply put a mark in the positive results and no mark in the negative results.

If you don't want to draw the grid lines, I've seen this done with either checks and x's or with open and filled circles.
posted by Netzapper at 3:41 PM on June 13

The data there doesn't seem complex and the data looks well-suited for a simple table to me. You just like the concept of moving to the right on a graph to convey time better than reading down a table? When there is no real useful aggregation for the range, graphs aren't really the right tool for the job. You could do a stacked bar chart I guess with a color for each test, but that's going to be more confusing than just handing out your tabular data.
posted by cmm at 3:48 PM on June 13

Please clarify if your data set is limited to the one subject you show or if you have a larger number of subjects of which the above is just one example.
posted by drpynchon at 3:49 PM on June 13

Do you mean that you have, for each case, 24 data points? And some reasonably large number of cases?

In that case, it would be simple to just have x= # treatments and y=cumulative proportion of cases resolved with that number of treatments. Then (if I understand your data right) you could either have three figures or one figure with three lines on it.

If you mean you have 24 data points all told, that table in your more inside should be all you need. Graphing that, if that's all you have, seems the equivalent of Calvin putting his report in a professional clear binder.
posted by ROU_Xenophobe at 3:52 PM on June 13

i think the table is probably the best best. drpynchon & ROU_Xenophobe: The data I listed (although unclear) is ALL the data. There is only one subject, with three tests, each measured 8 times. The more I think about it, the sillier it seems to use a graph. I'll do a table.

(unless some genius has a more creative way of displaying it--- i'm pretty sure this will be the best way to write it)

Thanks all!
posted by nitor at 3:59 PM on June 13

A Consumer Reports–style table with a table of filled-in or empty circles seems best to me.
posted by grouse at 4:17 PM on June 13

If you had lots of multiples of this data then I was going to suggest sparklines (like on page 18 here) but, yes, a table.
posted by primer_dimer at 1:57 AM on June 15