User talk:The Evil Spartan

The new algorithm isn't clear

It isn't clear what exact weights are older polls given. Please specify. Rami R 08:24, 2 October 2008 (UTC)

This was in response to a complaint I had on commons. It's a bit complex, and if it weren't housed on commons, I might consider it original research, but it is, so I'm not worried about that. It goes like this:

1. Full weight is given to the most recent 3 polls.
2. The algorithm checks if there were 86 polls in the past 15 days. If there were, it weights only these polls. For each day since the most recent poll, it will lose 1/15th of its weight. For example, if the most recent poll was August 27th, and there is a poll from August 15th, it will have only 1/5 (3/15) the weight.
3. If there are not that many polls in the past 15 days, it does the same as above, but weights according to the past 30 days. In this case, the August 15th poll will have a weight of 3/5 (18/30). This second provision is because some states are less polled than others, and aren't able to evaluate trends as well.

I can provide the OpenOffice Calc document I use if you would like. Is this clear enough? The Evil Spartan (talk) 15:17, 2 October 2008 (UTC)

Reasonably clear. 'though, this really should be written at the image description page. Otherwise you'll probably have users question your neutrality again. Rami R 15:28, 2 October 2008 (UTC)

Sure; do you know if commons accepts OpenOffice documents? I really hate to have made it so complex, but the previous method of just throwing everything together was leaving the maps hopelessly out of date; the IP user on my commons talk page had that correct. The Evil Spartan (talk)

Sorry for not replying earlier, RL stuff came up. I don't know if commons accepts OpenOffice. You'll need to ask there. Rami R 13:05, 5 October 2008 (UTC)

The new poll projection table

I love the new table you created, but there are editors who have questions about how the data was compiled with citations to back it up. Can you add your input in Talk:Statewide opinion polling for the United States presidential election, 2008? Steelbeard1 (talk) 22:18, 2 October 2008 (UTC)

Please include the old maps they are better than the 538 map. Thank you. —Preceding unsigned comment added by 152.23.33.228 (talk) 18:03, 8 October 2008 (UTC)

Arizona poll

You have marked the Zimmermann/Marketing Intelligence poll recently posted at Statewide opinion polling for the United States presidential election, 2008 as partisan, based − as it seems − on an article by the ”Arizona Daily Star“. However, the press release by the polling firms insists that “The survey was not commissioned by any political party or candidate.” and pollster.com describes the two pollsters as “non-partisan research and public relations firms”. I'm not sure which course to follow with this one, but I cannot see overwhelming proof of any bias. However, maybe we should just wait for the new Arizona poll by Rasmussen to be released soon. That should give us an accurate trend. FanofPolling (talk) 15:02, 27 October 2008 (UTC)

If you read the edit summary, I gave the reason: the poll was done by Democratic pollsters. The star specifically says "a poll commissioned by an organization affiliated with a specific candidate"; it exactly fits this definition. As such, you will notice fivethirtyeight did not mention it. The Evil Spartan (talk) 02:24, 28 October 2008 (UTC)