Interesting Article on Statistical Bias

[silverscreenselect]

Here's an interesting article (based on an excerpt from a current book) about how statistics and surveys can be biased even though there's no attempt by the author to do so. One example. Suppose 99 people conduct studies to find whether playing video games has a correlation with colon cancer, and they all show no correlation whatsoever. No medical journal is likely to publish these studies because they are not very interesting. Then the one hundredth study, which very likely is an outlier (a rare result that's significantly outside what's expected statistically), shows that playing video games does correlate with a lower rate of colon cancer. Of course, that's the study that's likely to get published (and subsequently reported in mainstream media) and "prove" a highly unlikely scientific theory.

https://medium.com/editors-picks/96eba8 ... cmpid=view

[Bob78164]

Here's an interesting article (based on an excerpt from a current book) about how statistics and surveys can be biased even though there's no attempt by the author to do so. One example. Suppose 99 people conduct studies to find whether playing video games has a correlation with colon cancer, and they all show no correlation whatsoever. No medical journal is likely to publish these studies because they are not very interesting. Then the one hundredth study, which very likely is an outlier (a rare result that's significantly outside what's expected statistically), shows that playing video games does correlate with a lower rate of colon cancer. Of course, that's the study that's likely to get published (and subsequently reported in mainstream media) and "prove" a highly unlikely scientific theory.

https://medium.com/editors-picks/96eba8 ... cmpid=view

Thanks. That was interesting. Now I'm off to find some purple pajamas. --Bob

[thguy65]