tag:blogger.com,1999:blog-15868947.post6191660588444135274..comments2023-10-29T06:41:23.910-07:00Comments on Halfway There: Roller-coaster pollingZenohttp://www.blogger.com/profile/09058127284297728552noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-15868947.post-61590806333938713532008-08-09T18:20:00.000-07:002008-08-09T18:20:00.000-07:00You're right, William. The talking heads never inc...You're right, William. The talking heads never include the specifics of what plus-or-minus 3 points really means. Even print media, where they have more room for details, avoid talking about confidence intervals and the <I>probability</I> of being within the so-called margin of error. It's a shortcut that misleads people and suggests more precision than is actually the case.<BR/><BR/>Don't hold your breath expecting them to fix that.Zenohttps://www.blogger.com/profile/09058127284297728552noreply@blogger.comtag:blogger.com,1999:blog-15868947.post-47917447696335792812008-08-09T14:06:00.000-07:002008-08-09T14:06:00.000-07:00Maybe I'm just not paying enough attention, but I ...Maybe I'm just not paying enough attention, but I never hear the talking heads say anything like "accuracy within 3 percentage points 90% of the time" -- it's always just "margin of error, 3 percentage points", implying that the error will <I>never</I> be outside that range.<BR/><BR/>Perhaps "margin of error" is more a term of art than I suspected?Williamhttps://www.blogger.com/profile/02052684196866992031noreply@blogger.comtag:blogger.com,1999:blog-15868947.post-52811194487272785362008-08-08T11:20:00.000-07:002008-08-08T11:20:00.000-07:00Would peckers account for holes in telephone polls...Would peckers account for holes in telephone polls?<BR/>Cuz if you pick a pack of polling peckers,<BR/>you'll probably find their fingers free<BR/>nearly never nine to nine,<BR/>when paleold voters dine in their homes at scheduled times,<BR/>while the working poor<BR/>ain't rolled in the door<BR/>from their second job at the store,<BR/>and the youngsters don't mind<BR/>not bein' landlined...<BR/>So the peckers<BR/>hardly penetrate,<BR/>so demoscriminate...<BR/>Just like a buncha peckers,<BR/>always skewing things up.<BR/><BR/>;-}Proud2bHumblehttps://www.blogger.com/profile/03691609895881003514noreply@blogger.comtag:blogger.com,1999:blog-15868947.post-87783178034074763032008-08-08T05:49:00.000-07:002008-08-08T05:49:00.000-07:00I hate that!They do it here, too. Even outside of ...I <I>hate</I> that!<BR/><BR/>They do it here, too. Even outside of election time. And don't get me started on referenda.<BR/><BR/>I wish I knew what to do to get a ban on polling. At least in the run-up to elections. Including bloody exitpolls.<BR/><BR/>/rant /rageJens Knudsen (Sili)https://www.blogger.com/profile/14078875730565068352noreply@blogger.comtag:blogger.com,1999:blog-15868947.post-20791139185870560652008-08-07T13:26:00.000-07:002008-08-07T13:26:00.000-07:00Idea: plot the result including the distribution. ...Idea: plot the result <I>including</I> the distribution. Of course, most people can't read a probability distribution worth a damn, so here's what I'm thinking...<BR/><BR/>Don't plot the curve. Plot a vertical bar shaded more or less densely according to the probability density function. Then for each [time-unit] of your tracking poll, plot a bar graph, but use these "density bars" instead of a simple height. Even better would be if you can smooth out the jumps at the edges of the bars.<BR/><BR/>Of course I can't program to save my life, so I don't have any examples of this... maybe a bleg...Anonymousnoreply@blogger.com