Comments for chatter.chaboud.com http://chatter.chaboud.com Striving to deliver the lowest signal-to-noise ratio on the internet... Tue, 25 Oct 2005 06:45:40 -0400 http://wordpress.org/?v=2.8.4 hourly 1 Comment on Updates to the site… by me http://chatter.chaboud.com/?p=122&cpage=1#comment-33 me Sat, 22 Jan 2005 15:42:55 +0000 /?p=122#comment-33 Apple IE doesn't even load the image. It also fails with susan.chaboud.com Apple IE doesn’t even load the image. It also fails with susan.chaboud.com

]]> Comment on Roll up your sleeves and sue… by Aaron http://chatter.chaboud.com/?p=123&cpage=1#comment-32 Aaron Sat, 22 Jan 2005 07:04:06 +0000 /?p=123#comment-32 I like the new look. I like the new look.

]]> Comment on Think that you can tell the difference between Chinese, Japanese, and Koreans? by Dan http://chatter.chaboud.com/?p=117&cpage=1#comment-31 Dan Wed, 31 Dec 1969 19:00:00 +0000 /?p=117#comment-31 6 out of 18. Not even average. 6 out of 18. Not even average.

]]> Comment on Think that you can tell the difference between Chinese, Japanese, and Koreans? by Arron http://chatter.chaboud.com/?p=117&cpage=1#comment-30 Arron Wed, 31 Dec 1969 19:00:00 +0000 /?p=117#comment-30 4 out of 18. I'm hopeless. 4 out of 18.

I’m hopeless.

]]> Comment on The importance of information control… by Daniel Bonniot http://chatter.chaboud.com/?p=113&cpage=1#comment-29 Daniel Bonniot Wed, 31 Dec 1969 19:00:00 +0000 /?p=113#comment-29 I observed the same thing, not only in Ohio but in other states and national exit polls too (I've kept a bunch of copies of the files). I've sent a message to CNN in case they have the time and will to clarify this issue, no answer yet. Below is what I wrote. Note that even with the most conservative rounding errors it cannot have been the 57 new respondents. It looks like they modified the weigths to make the polls match the results. At least they indicate the update time, so they're not trying to hide it. But given the doubt on the fairness of the counting due to the absence of trails, this is unsettling. Here's what I sent about the national polls to CNN (in the national polls the numbers for Kerry were 47% and 54% before the update, 45% and 52% afterwards). Hi, I've followed with much interest your coverage of these elections. However, one point about the exit polls troubled me. I noticed that their value evolved over time, with a slightly higher number of respondents, and a significant shift towards Georges W Bush. For instance, in the national exit polls: At 9.20 pm, there were 13407 respondents. Using conservative estimates of the rounding errors, this polls gave at least: 13407 * 0.46 *0.465 = 2867 males for Kerry 13407 * 0.54 * 0.535 = 3873 females for Kerry However at 1.36 am, the same poll had 13531 respondents and with the opposite estimate for rounding that gives only: 13531 * 0.46 * 0.455 = 2832 males for Kerry 13531 * 0.54 * 0.525 = 3836 females for Kerry That is, with additional respondents, Kerry has less suporters! I noticed similar changes in a least some state exit polls. I would be interested to understand: 1) why the polls were updated, and why more respondents were included 2) what explains the differences in the pourcentages between both versions. If I made no mistake, I have shown that it cannot just be the effect of the new respondents. If the answer to 2) is different weightings, could you explain how the weightings were modified, and what motivated this change? Thanks a lot for your answer. Best regards, Daniel Bonniot I observed the same thing, not only in Ohio but in other states and national exit polls too (I’ve kept a bunch of copies of the files). I’ve sent a message to CNN in case they have the time and will to clarify this issue, no answer yet. Below is what I wrote. Note that even with the most conservative rounding errors it cannot have been the 57 new respondents. It looks like they modified the weigths to make the polls match the results. At least they indicate the update time, so they’re not trying to hide it. But given the doubt on the fairness of the counting due to the absence of trails, this is unsettling.

Here’s what I sent about the national polls to CNN (in the national polls the numbers for Kerry were 47% and 54% before the update, 45% and 52% afterwards).

Hi,

I’ve followed with much interest your coverage of these elections. However, one point about the exit polls troubled me. I noticed that their value evolved over time, with a slightly higher number of respondents, and a significant shift towards Georges W Bush. For instance, in the national exit polls:

At 9.20 pm, there were 13407 respondents. Using conservative estimates of the rounding errors, this polls gave at least:
13407 * 0.46 *0.465 = 2867 males for Kerry
13407 * 0.54 * 0.535 = 3873 females for Kerry

However at 1.36 am, the same poll had 13531 respondents
and with the opposite estimate for rounding that gives only:
13531 * 0.46 * 0.455 = 2832 males for Kerry
13531 * 0.54 * 0.525 = 3836 females for Kerry

That is, with additional respondents, Kerry has less suporters!

I noticed similar changes in a least some state exit polls.

I would be interested to understand:

1) why the polls were updated, and why more respondents were included

2) what explains the differences in the pourcentages between both versions. If I made no mistake, I have shown that it cannot just be the effect of the new respondents.

If the answer to 2) is different weightings, could you explain how the weightings were modified, and what motivated this change?

Thanks a lot for your answer.

Best regards,

Daniel Bonniot

]]>