5 ms·

But here you go: >"the lung cancer mortality rate has risen substantially between 1914 and 1968 among persons who never smoked cigarettes. For white males the

by nonbel 8y ago

But here you go: >"the lung cancer mortality rate has risen substantially between 1914 and 1968 among persons who never smoked cigarettes. For white males the relative increase for ages 35--84 years has been about 15-fold; the relative increase for ages 65--84 years has been about 30-fold." https://www.ncbi.nlm.nih.gov/pubmed/285289 https://www.ncbi.nlm.nih.gov/pubmed/285289

cbkeller 8y ago
That's interesting -- doesn't seem to have a graph, but states a significant increase between 1914 and 1967 in nonsmokers. There doesn't appear to be any control for secondhand smoke, however, which could also be a significant factor over that time period.
- sanxiyn 8y ago
  My impression is that second-hand smoking is considered insufficient to explain epidemiology of non-smoker lung cancer; see my other comments.
- nonbel 8y ago
  Sure, another thing to consider is fallout incorporated into tobacco and then smoked... But do you see why I didn't care if a correlation existed? There are a bunch of ways to explain such a correlation. That line of reasoning was based on the affirming the consequent fallacy[1]: If H: Nuclear tests are a significant factor in lung cancer rate Then O: We should still see a correlation between lung cancer incidence and nuclear test frequency [in non-smokers] We can't conclude "O is true, therefore H" (since there are a million potential reasons for O besides H), so we don't learn anything from it. Now if O had been some very specific prediction unlikely to arise from other hypotheses, I would care. In terms of Bayes' Rule, I only care if p( O|H[0] ) dominates all the other possibilities (H[1:n]) in the denominator. The mere existence of a positive correlation isn't specific enough for me to care: p( H[0]|O ) = p( H[0] )*p( O|H[0] )/sum( p( H[0:n] )*p( O|H[0:n] ) ) [1] https://en.wikipedia.org/wiki/Affirming_the_consequent https://en.wikipedia.org/wiki/Affirming_the_consequent
  - sanxiyn 8y ago
    Yes, epidemiology is not deductive, and in the end it depends on whether observation is "specific enough" for you st unlikely to arise from other hypotheses, which in turn depends on your prior, which is informed by laboratory results in addition to field results. Epidemiology is a truly fascinating subject.
    - nonbel 8y ago
      > "epidemiology is not deductive" -Not sure what you mean by this, it is quite possible to deduce precise predictions from an epidemiological model. See, eg, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2007940/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2007940/ >"it depends on whether observation is "specific enough" for you [is] unlikely to arise from other hypotheses, which in turn depends on your prior" -This has to do with the likelihoods, p( O|H[i] ), not the priors: p( H[i] ). >"[the prior] is informed by laboratory results in addition to field results" - Yes
      - sanxiyn 8y ago
        Affirming the consequent is a deductive fallacy. It is not a fallacy in probabilistic reasoning, because the consequent is an evidence of the antecedent. Re: laboratory results, it is noteworthy smoking and non-smoking lung cancer have different molecular signature. Lung cancer probably should be considered two separate diseases, and this probably overstates smoking connection because some lung cancer by smokers is of non-smoking kind and not caused by smoking. This can be better understood by including such signatures in epidemiological study.
        nonbel 8y ago
        >"Affirming the consequent is a deductive fallacy. It is not a fallacy in probabilistic reasoning" Nevertheless, people commit that fallacy all the time in epidemiology. They say stuff like "if smoking caused cancer, then people who smoked would get cancer more often". Then if there is a (statistically significant) positive correlation between cancer and smoking, they conclude smoking causes cancer. That is quite obviously affirming the consequent. I explain the (completely different and correct) probabilistic way to think about this situation using Bayes' rule in my post. It has nothing to do with arbitrary cut-offs for telling us whether a correlation exists or not. Instead it requires comparing the relative performance of the various explanations (at least the top few candidates since extremely unlikely ones can be dropped from the sum in the denominator).
sanxiyn 8y ago
Yup, epidemiology of non-smoker lung cancer is an active area of research. (Epidemiology of smoker lung cancer is not; it's very well established.) I haven't seen this old article, but it sounds perfectly right; the quoted result is probably due to diagnostic criteria. Lung cancer in never smokers (2009) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3170525/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3170525/ shows the field in confusion. Radon, indoor air pollution, second-hand smoking are named as possible sources, but it's still not satisfactory.

by nonbel 8y ago

cbkeller 8y ago

That's interesting -- doesn't seem to have a graph, but states a significant increase between 1914 and 1967 in nonsmokers. There doesn't appear to be any control for secondhand smoke, however, which could also be a significant factor over that time period.

sanxiyn 8y ago
My impression is that second-hand smoking is considered insufficient to explain epidemiology of non-smoker lung cancer; see my other comments.
nonbel 8y ago
Sure, another thing to consider is fallout incorporated into tobacco and then smoked... But do you see why I didn't care if a correlation existed? There are a bunch of ways to explain such a correlation. That line of reasoning was based on the affirming the consequent fallacy[1]: If H: Nuclear tests are a significant factor in lung cancer rate Then O: We should still see a correlation between lung cancer incidence and nuclear test frequency [in non-smokers] We can't conclude "O is true, therefore H" (since there are a million potential reasons for O besides H), so we don't learn anything from it. Now if O had been some very specific prediction unlikely to arise from other hypotheses, I would care. In terms of Bayes' Rule, I only care if p( O|H[0] ) dominates all the other possibilities (H[1:n]) in the denominator. The mere existence of a positive correlation isn't specific enough for me to care: p( H[0]|O ) = p( H[0] )*p( O|H[0] )/sum( p( H[0:n] )*p( O|H[0:n] ) ) [1] https://en.wikipedia.org/wiki/Affirming_the_consequent https://en.wikipedia.org/wiki/Affirming_the_consequent
- sanxiyn 8y ago
  Yes, epidemiology is not deductive, and in the end it depends on whether observation is "specific enough" for you st unlikely to arise from other hypotheses, which in turn depends on your prior, which is informed by laboratory results in addition to field results. Epidemiology is a truly fascinating subject.
  - nonbel 8y ago
    > "epidemiology is not deductive" -Not sure what you mean by this, it is quite possible to deduce precise predictions from an epidemiological model. See, eg, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2007940/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2007940/ >"it depends on whether observation is "specific enough" for you [is] unlikely to arise from other hypotheses, which in turn depends on your prior" -This has to do with the likelihoods, p( O|H[i] ), not the priors: p( H[i] ). >"[the prior] is informed by laboratory results in addition to field results" - Yes
    - sanxiyn 8y ago
      Affirming the consequent is a deductive fallacy. It is not a fallacy in probabilistic reasoning, because the consequent is an evidence of the antecedent. Re: laboratory results, it is noteworthy smoking and non-smoking lung cancer have different molecular signature. Lung cancer probably should be considered two separate diseases, and this probably overstates smoking connection because some lung cancer by smokers is of non-smoking kind and not caused by smoking. This can be better understood by including such signatures in epidemiological study.
      - nonbel 8y ago
        >"Affirming the consequent is a deductive fallacy. It is not a fallacy in probabilistic reasoning" Nevertheless, people commit that fallacy all the time in epidemiology. They say stuff like "if smoking caused cancer, then people who smoked would get cancer more often". Then if there is a (statistically significant) positive correlation between cancer and smoking, they conclude smoking causes cancer. That is quite obviously affirming the consequent. I explain the (completely different and correct) probabilistic way to think about this situation using Bayes' rule in my post. It has nothing to do with arbitrary cut-offs for telling us whether a correlation exists or not. Instead it requires comparing the relative performance of the various explanations (at least the top few candidates since extremely unlikely ones can be dropped from the sum in the denominator).

sanxiyn 8y ago

Yup, epidemiology of non-smoker lung cancer is an active area of research. (Epidemiology of smoker lung cancer is not; it's very well established.) I haven't seen this old article, but it sounds perfectly right; the quoted result is probably due to diagnostic criteria. Lung cancer in never smokers (2009) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3170525/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3170525/ shows the field in confusion. Radon, indoor air pollution, second-hand smoking are named as possible sources, but it's still not satisfactory.