.comment-link {margin-left:.6em;}

Born at the Crest of the Empire

Saturday, May 16, 2009

Picture of the Day - Bad data point

The right is crowing about a new Gallup poll showing that America is now "pro-life."

I'm not a statistical expert, but it's my hunch that they just pulled a non-representative sample. (sample size 971.)


I mean, after ten years of unmoving support at 47-50, the graph suddenly completely inverts at a time when Republicans and Republican issues are at their very lowest ebb?

Of course, what else do they have to crow about?

2 Comments:

  • I'm not an expert either, but I know enough about statistics that with that sample size, they have a 95% chance of being within about 3% plus or minus.

    Unless they purposely pulled a bad sample, you can be pretty certain of the results.

    The question asked is important too, but I'm guessing if they are showing historical benchmarks, the question has to be the same.

    By Blogger Praguetwin, at 1:59 PM  

  • I think it is.

    I'm guessing bad sample simply because I have no other supporting evidence or reason to support that result.

    If it were a normal candidate poll, I would just say, "let's wait for one or two others to prove or disprove the result," but since it's an issue item outside of a peak interest period, we may have to wait.

    (On the other hand, interest in this result might get another organization to add the question to their normal set approval/state of the nation polling.)

    Again, not saying they're cheating or slanting, just that they probably polled a bad result (through proper methodology.) It happens.

    Let's wait and see if we get some more polling on the topic.

    By Blogger mikevotes, at 2:54 PM  

Post a Comment

<< Home