Debunked: 14,000 Fukushima Deaths in U.S.

SpunkyMonkey

Last Monday a press release announced the shocking result of a new study:

Medical Journal Article: 14,000 U.S. Deaths Tied to Fukushima Reactor Disaster Fallout


Immediately seeing major problems with that study by Mangano & Sherman (M&S), I asked a statistician what he thought of it. He crunched the data and while he found several devastating statistical problems, his most remarkable finding was that the U.S. infant-death data M&S report as being from the CDC does not jibe with the actual CDC infant-death data for the same weeks.

The M&S infant-death data allegedly from the CDC can be seen here (go to Table 3, page 55). And the actual CDC infant-death data can be seen here (go to Locations, scroll down and select Total and press Submit for the data; the data for infants is in the Age column entitled "Less than 1"). The mismatching data sets are included at the end of this post, and with the links I've provided here, everything I'm saying can be independently confirmed by the reader.

Here are the mismatching data sets, note that post-Fuku weeks 15 through 24 do match:

2010 (weeks 50-52) 2011 (weeks 1-25)

wk M&S CDC
50 : 202 216
51 : 129 143
52 : 113 130
1 : 158 183
2 : 177 208
3 : 158 185
4 : 148 171
5 : 178 208
6 : 173 182
7 : 188 206
8 : 158 186
9 : 174 199
10 : 165 182
11 : 188 209
12 : 201 211
13 : 210 213
14 : 198 204
15 : 163 163
16 : 188 188
17 : 200 200
18 : 196 196
19 : 214 214
20 : 224 224
21 : 196 196
22 : 152 152
23 : 174 174
24 : 191 191
25 : 215 217

The nature of the mismatch is that all the pre-Fukushima M&S data points are lower than the actual CDC data points and bias the data set to to a statistically significant increase in post-Fukushima infant deaths. But in the actual CDC data, there is no statistically significant increase. The statistician also found that even M&S's data for all-age deaths was in fact not statistically significant, contrary to the claim of M&S.

Why the infant data are mismatched is not understood at this time. However, a review of the archived copies of the Morbidity and Mortality Weekly Report archive finds that the historically released data points for the weeks in question jibe with the CDC's MMWR database. So I see no reason to believe the CDC's online data are not the true data.

rootX

It's too speculative and resources you linked don't seem credible enough.

Here's one response to the OP article:
http://nuclearpoweryesplease.org/blo...oseph-mangano/
I didn't read through the article.

SpunkyMonkey

I'm sorry you don't find the CDC to be credible enough. And my only other links were to the study in question and to its press release. Given that the study is the topic of discussion, linking to it directly is far more credible than linking to a opinion piece on it (and the piece you link to I agree with, but it's on an earlier version of the study). Your comment is nonsense! :)

rootX

To make it clear, I was only referring to the the Mangano & Sherman actual study. I wasn't questing the CDC data. I only read the "14,000 Fukushima Deaths in U.S." in the title and Mangano & Sherman from the OP and found it too ridiculous that I didn't care to pay attention to the rest of the post. I don't know why someone even need to take the Mangano & Sherman study seriously.

Pengwuino

The CDC did not write that article, they simply used CDC data and misrepresented it. The journal they "published" to is not a legitimate journal either as far as we can tell.

SpunkyMonkey

Um, I linked to the CDC database and rootX said the "resources you linked don't seem credible enough." So becasue I linked to the CDC database you to believe I thought the CDC wrote the study, oy vey!

Jimmy Snyder

I couldn't find any independent review of the journal although I have to admit I only looked at two pages of google results. According to their own website (not that I would take their word for anything now) says that they have been publishing since 1971. Here are a couple of quotes from their website.
I question the quality of their peer referee process. And they might rank in the bottom 5 for all I know. Have you got some reason to believe they are not what they say they are?

SpunkyMonkey

There's some review of the journal in this critical blog.

Jimmy Snyder

It sounds like a journal that was living near the edge of the cliff and just fell off of it.

D H

Yet another critique: http://blogs.scientificamerican.com/...a-death-study/

SpunkyMonkey

This explains their data manipulation that resulted in the mismatch above. From page 52 of their paper (bold emphasizes the manipulation):


So Mangano & Sherman's pre-Fuku data (weeks 50 - 11) only included 104 cities (of the 122 cities in the CDC database), but for the post-Fuku data (weeks 12 - 25) they included 119 cities. The exclusion of cities from the pre-Fuku data resulted in a lowering of total-death counts that thereby produced a statistically significant post-Fuku increase in deaths. The statistician I'm communicating with eliminated the same cities from the post-Fuku data so that the number of cities remains constant at 104 across the data set, and then the statistical significance disappears.

A crystal clear example of massaging the data until it fits your thesis -- pathetic!

Antiphon

Too bad research fraud doesn't carry civil peanalties.

If I sell a can of dog food with the wrong order of ingredients I can loose my license to manufacture pet food.

Astronuc

In some cases, it could, and in some cases, fraud is a criminal offense. If one's work is funded, then the funding institution could sue one to recover the funding. On the other hand, if one's fraudulent research promotes an agenda of the funding source, then one wouldn't risk a lawsuit.

In many pure scientific cases, scientific conduct might get one fired from job, e.g., at a university.

Of course, these days, I'd expect faulty research to be covered by freedom of speech, much the way that faulty or fraudulent financial information, e.g., AAA ratings on junk financial instruments was considered an opinion covered by freedom of speech.

Vanadium 50

Civil? It should be criminal. Andrew Wakefield, for example, should be tried for homicide for every measles-related death caused by his fraudulent (and bought and paid for) study.

Ideally in Texas.

AlephZero

Hmm ... http://www.scimagojr.com/journalrank.php?category=2719 seems to rank it #55 out of 120, not #5.

But I could believe that 14,000 new-age airheads went into a state of blind panic and starved their kids to death for fear of feeding them something dangerous...

rootX

I also find it disturbing that someone would actually report on this study:
http://www.prnewswire.com/news-relea...135859288.html (from OP)
Whoever wrote the news article seems to be consulted only two people, Joseph Mangano and Janette Sherman.

That could have been a valid study to believe in

Yes, I agree that my comment was nonsense. It took me few hours to realize that you were actually trying to debunk a study that I dismissed just looking at the title

D H

Also from the paper,

The gap in changes for infant deaths (+1.80% in the latter 14 weeks, –8.37% for the earlier 14 weeks) was even larger.

If a 1.80% deviation is statistically significant, what does that make a deviation that is over four times larger? This must prove that those babies born before Fukushima were somehow prescient and knew that they had to stay alive to offset the post-Fukushima deaths.

Or maybe it means that the shorter a time interval one looks at, the larger the deviation. Or it could just mean that the authors cooked the books.