tag:blogger.com,1999:blog-6007032115980019186.post2425523062684724891..comments2023-11-02T12:21:34.000+11:00Comments on Mark the Ballot: Polling updateMark Graphhttp://www.blogger.com/profile/10462713733051104779noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-6007032115980019186.post-35193406546271763692019-05-13T23:08:54.069+10:002019-05-13T23:08:54.069+10:00Definitely fewer polls and fewer polling companies...Definitely fewer polls and fewer polling companies. But this should not cause herding<br />Mark Graphhttps://www.blogger.com/profile/10462713733051104779noreply@blogger.comtag:blogger.com,1999:blog-6007032115980019186.post-89774215270991188972019-05-13T22:36:21.610+10:002019-05-13T22:36:21.610+10:00The Essential Report used to constantly disclose t...The Essential Report used to constantly disclose that they were using a rolling 2 week average for their polling, for example:<br />https://www.essentialvision.com.au/federal-politics-%e2%80%93-voting-intention-127<br /><br />"The data in the above tables comprise 2-week averages derived from the first preference/leaning to voting questions."Anonymoushttps://www.blogger.com/profile/12513414072917029876noreply@blogger.comtag:blogger.com,1999:blog-6007032115980019186.post-83030384603238236132019-05-13T19:42:22.364+10:002019-05-13T19:42:22.364+10:00There is another possibility. Building the sample...There is another possibility. Building the sample from an aggregate of sub-samples, where each sub-sample is taken from a division of the population, and the size of each sub-sample is in the same proportion to the total sample size that the sub-population bears to the total population.<br /><br />One way they could do this is on voter occupation. So if teachers are 3% of the population, they select 3% of the sample to be teachers.<br /><br />This would reduce the variance because occupation is strongly correlated with voting. The population mean of near 50% results in a high variance (i.e. var = npq). But in some occupations, you can get 75% of the population voting the same way - and when such a sub population is sampled, the sub-sample variance is reduced because we are not near 50%.<br />Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6007032115980019186.post-14946067127765971922019-05-13T10:43:24.688+10:002019-05-13T10:43:24.688+10:00Mark,my perception is there are less polls this ti...Mark,my perception is there are less polls this time. for instance we have only Newspoll one week out from the election. I cannot remember that.<br /><br />Could less polls be a possibility for the 'herding'Not Trampishttps://www.blogger.com/profile/12738633092867411422noreply@blogger.com