She gave a TED talk on this, too:

http://tedxtalks.ted.com/video/Why-data-should-be-over-the-cou

"Sadly, NASA had data that could have supported the argument not to launch. That data, however, was not displayed clearly during the debate over whether to proceed. Data design experts have since redrawn NASA’s graph depicting the test launches’ incidents of O-ring thermal distress, arranged by launch temperature. As the Report of the Presidential Commission on the Space Shuttle Challenger Accident (the Rogers Report) demonstrated, the chart used the night before the disaster depicted data from failed launches. This graph offered no insight into how the O-rings functioned because it failed to include data from the 17 successful launches.

Viewing the original chart, which shows the outside temperature during launches that had O-ring incidents, nothing jumps out to indicate the January launch would fail. The improved chart, which includes data when O-rings functioned property, tells another story."

This thinking is similar to a stash with statistician Abraham Wald reinforcing WWll British bombers. 

Edward Tufte tells a similar story about PowerPoint ruining data prior to the Columbia crash.

http://www.edwardtufte.com/bboard/q-and-a-fetch-msg?msg_id=0000OR