4 ms·

Oh geez. From the New York Times article: "In most respects, there was no significant difference between the Kindle readers and the paper readers: the emotiona

by cwyers 12y ago

Oh geez. From the New York Times article: "In most respects, there was no significant difference between the Kindle readers and the paper readers: the emotional measures were roughly the same, and both groups of readers responded almost equally to questions dealing with the setting of the story, the characters and other plot details. But, the Kindle readers scored significantly lower on questions about when events in the story occurred. They also performed almost twice as poorly when asked to arrange 14 plot points in the correct sequence." So, using a battery of test questions divided into five subjects, INCLUDING TIME AND EVENTS, there was no significant difference between the two groups. But on an entirely different test methodology, the Kindle readers scored significantly worse than the paper readers on plot orders. So they go, "Well, we have a battery of results that shows no difference, and one result that shows a difference. Let's backfit an explanation to the one interesting result!" It's that XKCD strip about jelly beans.

yasth 12y ago
It's that XKCD strip about jelly beans. ( http://www.xkcd.com/882/ http://www.xkcd.com/882/ )
tzs 12y ago
> So, using a battery of test questions divided into five subjects, INCLUDING TIME AND EVENTS, there was no significant difference between the two groups It says right in the part you quoted that the Kindle readers scored significantly lower on the event time questions. They also did poorly on the separate plot ordering test. I fail to see any problem with this, since I would expect the to be good correlation between how well people do on questions about event timing in the story and how well they can order a timeline of the plot.
- sixbrx 12y ago
  I think parent is pointing out that there could be a problem of "mutliple testing" in the statistics used. If you do enough tests for statistical significance, and do not correct for multiple testing, then some fraction (your cutoff, typically .05) will be expected to test positive even when nothing is happening (null hypothesis holds).

by cwyers 12y ago

yasth 12y ago

It's that XKCD strip about jelly beans. ( http://www.xkcd.com/882/ http://www.xkcd.com/882/ )

tzs 12y ago

> So, using a battery of test questions divided into five subjects, INCLUDING TIME AND EVENTS, there was no significant difference between the two groups It says right in the part you quoted that the Kindle readers scored significantly lower on the event time questions. They also did poorly on the separate plot ordering test. I fail to see any problem with this, since I would expect the to be good correlation between how well people do on questions about event timing in the story and how well they can order a timeline of the plot.

sixbrx 12y ago
I think parent is pointing out that there could be a problem of "mutliple testing" in the statistics used. If you do enough tests for statistical significance, and do not correct for multiple testing, then some fraction (your cutoff, typically .05) will be expected to test positive even when nothing is happening (null hypothesis holds).