4 ms·

Sure sounds like the Clustering Illusion[1] to me. Just another Erin Brockovich-style failure[2] to understand statistics. [1] http://www.skepdic.com/clusterin

by cjensen 12y ago

Sure sounds like the Clustering Illusion[1] to me. Just another Erin Brockovich-style failure[2] to understand statistics. [1] http://www.skepdic.com/clustering.html http://www.skepdic.com/clustering.html [2] http://www.slate.com/articles/health_and_science/medical_examiner/2013/03/cancer_cluster_in_toms_river_new_jersey_the_link_to_a_superfund_site_is.html http://www.slate.com/articles/health_and_science/medical_exa...

mpdehaan2 12y ago
The possible break in this is: "Since then, Griffin has compiled a list of 38 American soccer players -- 34 of them goalies". This at least seems to correlate something goalie related, provided Griffin was not drawing from a data-pool that was biased towards goalies. It might correlate something else that goalies do, or some mental desire to be a goalie with a weird genetic variant, and so on, but it seems interesting that the position is a factor.
- cjensen 12y ago
  Griffin is, perhaps without intending to, doing exactly what you hope she isn't: she's making a list of goalies in the US who have cancer. Who else is going to contact her to get their name on the list? The act of collecting a list is an error since the data collected is entirely meaningless. Here's how she should be doing it: randomly select a few cities in the US, then check the cancer rate among soccer players in those cities.
Retric 12y ago
There are lot's of known cancer clusters because disease clusters refers to groupings not just by location. It's a little more complex than what you linked to as the background cancer rate varies on a regional basis. But more importantly Cancer is a group of more than 100 different diseases. If an area get's 10% more cancer than expected but twice the breast cancer then there is probably a local cause. But, again the statistics get's far more complex than just X expected rate of any type of cancer vs X observed rate. So, "Blood cancers like lymphoma and leukemia dominate the list." is by far the most important part. And more importantly there is lot's of precedence for occupational links to cancer mesotheliomas being the most well known. http://en.wikipedia.org/wiki/Mesothelioma http://en.wikipedia.org/wiki/Mesothelioma Though Tobacco is responsible for far more total cancer than asbestos the occupational links are less well known.
- cjensen 12y ago
  "If an area get's 10% more cancer than expected but twice the breast cancer then there is probably a local cause." That is incorrect: deviation percentage is meaningless without a sample size. In a group of 100 people, twice the rate of cancer means nothing. In a group of 1 Billion people, 10% is significant. You're right that blood cancer is the most important part of the article... those are the cancers which are most common in children and therefore the cancers most likely to be found in an coincidental cluster in children.
  - Retric 12y ago
    10% for even just 1 million people is significant... Wait, let's look at a map of US cancer rates: http://www.cdc.gov/cancer/dcpc/data/state.htm http://www.cdc.gov/cancer/dcpc/data/state.htm Hmm, yep that's plenty of data and a wide enough gap to support regional differences. Yet, there are reasons you don't hear state X has higher incidents of cancer. Edit: Sorry for the flippant response, but no shit sample size matters.
    - cjensen 12y ago
      No worries. My point is really to say, "what kind of sample size do you have in mind"? In this instance, the sample size seems to be everyone in a particular region who has played soccer. That's just too vague, in my opinion, to attempt to draw a conclusion.
hudibras 12y ago
Griffin collected names through personal experience with sick players, and acknowledges that her list is not a scientific data set. But it’s enough to make her ask whether crumb rubber artificial turf...is safe for the athletes and kids who play on it. That's all she's saying; doesn't seem unreasonable to me. Also, the Erin-Brockovich-was-wrong story is overly simplistic. Many epidemiology studies are inconclusive because of small sample sizes. What is known is that the chromium levels in some neighborhoods in Hinkley, California were hundreds or thousands of times the normal concentration, PG&E settled with the affected families for a nine-digit sum, and that the most-affected neighborhoods have been razed. http://www.motherjones.com/environment/2013/05/erin-brockovich-hinkley-california-junk-science http://www.motherjones.com/environment/2013/05/erin-brockovi...
- cjensen 12y ago
  Settling just means that PG&E did a risk analysis on how they thought a jury might respond. Yes, the Chromium was high, but high-Chromium combined with a cancer cluster is still just a reasonable outcome from random probability.

by cjensen 12y ago

mpdehaan2 12y ago

The possible break in this is: "Since then, Griffin has compiled a list of 38 American soccer players -- 34 of them goalies". This at least seems to correlate something goalie related, provided Griffin was not drawing from a data-pool that was biased towards goalies. It might correlate something else that goalies do, or some mental desire to be a goalie with a weird genetic variant, and so on, but it seems interesting that the position is a factor.

cjensen 12y ago
Griffin is, perhaps without intending to, doing exactly what you hope she isn't: she's making a list of goalies in the US who have cancer. Who else is going to contact her to get their name on the list? The act of collecting a list is an error since the data collected is entirely meaningless. Here's how she should be doing it: randomly select a few cities in the US, then check the cancer rate among soccer players in those cities.

Retric 12y ago

There are lot's of known cancer clusters because disease clusters refers to groupings not just by location. It's a little more complex than what you linked to as the background cancer rate varies on a regional basis. But more importantly Cancer is a group of more than 100 different diseases. If an area get's 10% more cancer than expected but twice the breast cancer then there is probably a local cause. But, again the statistics get's far more complex than just X expected rate of any type of cancer vs X observed rate. So, "Blood cancers like lymphoma and leukemia dominate the list." is by far the most important part. And more importantly there is lot's of precedence for occupational links to cancer mesotheliomas being the most well known. http://en.wikipedia.org/wiki/Mesothelioma http://en.wikipedia.org/wiki/Mesothelioma Though Tobacco is responsible for far more total cancer than asbestos the occupational links are less well known.

cjensen 12y ago
"If an area get's 10% more cancer than expected but twice the breast cancer then there is probably a local cause." That is incorrect: deviation percentage is meaningless without a sample size. In a group of 100 people, twice the rate of cancer means nothing. In a group of 1 Billion people, 10% is significant. You're right that blood cancer is the most important part of the article... those are the cancers which are most common in children and therefore the cancers most likely to be found in an coincidental cluster in children.
- Retric 12y ago
  10% for even just 1 million people is significant... Wait, let's look at a map of US cancer rates: http://www.cdc.gov/cancer/dcpc/data/state.htm http://www.cdc.gov/cancer/dcpc/data/state.htm Hmm, yep that's plenty of data and a wide enough gap to support regional differences. Yet, there are reasons you don't hear state X has higher incidents of cancer. Edit: Sorry for the flippant response, but no shit sample size matters.
  - cjensen 12y ago
    No worries. My point is really to say, "what kind of sample size do you have in mind"? In this instance, the sample size seems to be everyone in a particular region who has played soccer. That's just too vague, in my opinion, to attempt to draw a conclusion.

hudibras 12y ago

Griffin collected names through personal experience with sick players, and acknowledges that her list is not a scientific data set. But it’s enough to make her ask whether crumb rubber artificial turf...is safe for the athletes and kids who play on it. That's all she's saying; doesn't seem unreasonable to me. Also, the Erin-Brockovich-was-wrong story is overly simplistic. Many epidemiology studies are inconclusive because of small sample sizes. What is known is that the chromium levels in some neighborhoods in Hinkley, California were hundreds or thousands of times the normal concentration, PG&E settled with the affected families for a nine-digit sum, and that the most-affected neighborhoods have been razed. http://www.motherjones.com/environment/2013/05/erin-brockovich-hinkley-california-junk-science http://www.motherjones.com/environment/2013/05/erin-brockovi...

cjensen 12y ago
Settling just means that PG&E did a risk analysis on how they thought a jury might respond. Yes, the Chromium was high, but high-Chromium combined with a cancer cluster is still just a reasonable outcome from random probability.