Estimated 655,000 dead in Iraq

[kallend 2,148]

Quote

First, I made an error in an earlier statement.

Change From:
Percent of Houses Surveyed in Iraq: 0.00047

Change To:
Percent of Houses Surveyed in Iraq: 0.047

Second, the sample size does contribute to the margin of error. I consider the margin of error from 392,979 to 942,636 to be large and not acceptable for a statistic used to affect social policy.

Yes, you made an error.

And this isn't social policy, it's foreign policy.

And even at the lower level of the interval, it's still way more than the "official" figures.

...

The only sure way to survive a canopy collision is not to have one.

[Butters 0]

Quote

Quote

First, I made an error in an earlier statement.

Change From:
Percent of Houses Surveyed in Iraq: 0.00047

Change To:
Percent of Houses Surveyed in Iraq: 0.047

Second, the sample size does contribute to the margin of error. I consider the margin of error from 392,979 to 942,636 to be large and not acceptable for a statistic used to affect social policy.

Yes, you made an error.

And this isn't social policy, it's foreign policy.

And even at the lower level of the interval, it's still way more than the "official" figures.

You can change previous statements from social to foreign. I agree that the lower level of the interval is way more than the "official" figures. However, this does not make either of the figures correct.

"That looks dangerous." Leopold Stotch

[jcd11235 0]

Quote

Quote

Quote

I understand sampling theory.
.

Apparently you don't understand it very well. The sample size was quite adequate for an acceptable margin of error. The most likely source of error is non-randomness of the sample, not its size.

Gotta agree, and the survey question and analysis itself is likely more in question.

If someone came to my work and asked 100 people how many skydivers they know, the total would be about 75. And all 75 of those would be referring to me. That one guy in 1500 or so, and maybe some one-timers that did tandems once. The surveyer wouldn't know the difference. But multiple counting the same guy would be the big difference. "Clusters" of homes indeed.

Likely the survey was written poorly and the results purposely analyzed incorrectly to make the results dramatic. It helps to get published, you get more grant funding.

Another reason that it's nice to validate inferences with testing and independent studies on other factors.

Nothing wrong with the sample size though - the CI is very large, but it does bound the lower end at a significant number and establishes an error range. So what?

I haven't read this particular study in its entirety, but when NPR was talking about it, they mentioned that this was the only death count that involved looking at death certificates.

Somehow I doubt the results of a Johns Hopkins affiliated study/survey used were purposely analyzed incorrectly.

I remember another study a couple years ago that put the Iraqi death toll at well over 100,000, so the conclusions of this study do not seem unreasonable.

Math tutoring available. Only $6! per hour! First lesson: Factorials!