3 ms·

Just this comment again: https://news.ycombinator.com/item?id=31158391 https://news.ycombinator.com/item?id=31158391 :-(

by dash2 5y ago

artemonster 5y ago
At this point the famous „correlation is not causaution“ quote is so abused that its on par with „goto is bad“. Studying correlation does not dismiss the research itself
- dash2 5y ago
  But here it's really important, because people might mistakenly believe that this research proves that (a) you shouldn't wear a mask (b) you shouldn't tell people to wear masks.
ltbarcly3 5y ago
No, that comment does not apply here at all. The saying is "Correlation does not imply Causation". This saying is correct, you can have correlation for lots of reasons without a causal relationship. However, causation does imply correlation. You cannot have causation without correlation. The contra-positive of this is no-correlation implies no-causation. This study shows a lack of correlation in the data that they have, which does, within the limitations of that data, imply a lack of causation. There could be confounding factors here, but it isn't completely irrational to think that wearing a cotton mask that only filters out about 10% of the particles in question (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7185834/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7185834/) might not be very effective at preventing transmission. If not, it is important that governments start stockpiling effective masks to prepare for future virus outbreaks.
- dash2 5y ago
  You are mistaken. You can have causation without correlation. For example, suppose my umbrella keeps me dry. But I only take my umbrella out when it's raining. The correlation between my taking my umbrella and my being wet is zero: I either don't get rained on because I have an umbrella, or because it isn't raining; I'm only wet from other causes, unrelated to umbrellas or rain. However, my umbrella causes me to be dry. If I didn't take my umbrella out, holding everything else constant, I would get wet. Here are more details and examples: https://theincidentaleconomist.com/wordpress/causation-without-correlation-is-possible/ https://theincidentaleconomist.com/wordpress/causation-witho... In this case, it's quite possible that (a) mask-wearing keeps you safe but (b) mask-wearing is only brought in when Covid is bad; and that the two countervailing effects lead to an estimated correlation of zero.
  - ltbarcly3 5y ago
    Yes, if you very carefully set up the data to avoid one entire class of data then you don't have any examples of that class. When you say `mask-wearing is only brought in when Covid is bad;` that is a hypothesis that can be tested. If that is true then data rolled up at the country level would not be sufficient to determine whether masks are effective, this is a confounding effect. That is a lack of ability to find correlation in the data. The correlation is still there, your data is just biased in a way that cancels it out or has entire classes of scenarios missing. My point is that umbrellas ARE correlated with staying dry in the rain, otherwise we shouldn't bother with them as they don't work. The data collected in your example is just extremely biased. The data in this study could be biased, and I call that out. However, at a logical level, you MUST have correlation if there is causation, that is very very nearly a tautology. What does it mean to cause something if the outcome is independent?
    - dash2 5y ago
      My example wasn't set up to avoid one class of data. I said that I might get wet for other, non-rain-related reasons. So there is data with me being wet and with me being dry. And in the data, the observed correlation is zero, even though there is causation. That is because there is a confound. Also, in my toy example, the data collection process is not biased. You can collect the entire population of my days out, and you will still find no correlation. The problem is not sampling bias, it is an unobserved confound. That was just a toy example. But the linked paper may be an important real world example! You're right that "mask-wearing is only brought in when Covid is bad" is testable. Does the paper test it? Does it attempt to control for endogeneity of mask-wearing? If not, then it may be reporting a spurious null result. You're right that causation implies correlation in the following sense. "X causes Y" means that if we hold everything else equal, and change X, then Y changes. But that requires holding everything else equal. It's entirely possible that in the data we observe, we see no correlation, because an unobserved confound is biasing our estimates towards 0. That is why serious social scientists spend so much time looking for random variation, and worrying about unobserved confounds. In particular, it is just false to say "they found no correlation, so that proves no causation".

Just this comment again: https://news.ycombinator.com/item?id=31158391 https://news.ycombinator.com/item?id=31158391 :-(

by dash2 5y ago

Just this comment again: https://news.ycombinator.com/item?id=31158391 https://news.ycombinator.com/item?id=31158391 :-(

artemonster 5y ago

At this point the famous „correlation is not causaution“ quote is so abused that its on par with „goto is bad“. Studying correlation does not dismiss the research itself

dash2 5y ago
But here it's really important, because people might mistakenly believe that this research proves that (a) you shouldn't wear a mask (b) you shouldn't tell people to wear masks.

ltbarcly3 5y ago

No, that comment does not apply here at all. The saying is "Correlation does not imply Causation". This saying is correct, you can have correlation for lots of reasons without a causal relationship. However, causation does imply correlation. You cannot have causation without correlation. The contra-positive of this is no-correlation implies no-causation. This study shows a lack of correlation in the data that they have, which does, within the limitations of that data, imply a lack of causation. There could be confounding factors here, but it isn't completely irrational to think that wearing a cotton mask that only filters out about 10% of the particles in question (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7185834/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7185834/) might not be very effective at preventing transmission. If not, it is important that governments start stockpiling effective masks to prepare for future virus outbreaks.

dash2 5y ago
You are mistaken. You can have causation without correlation. For example, suppose my umbrella keeps me dry. But I only take my umbrella out when it's raining. The correlation between my taking my umbrella and my being wet is zero: I either don't get rained on because I have an umbrella, or because it isn't raining; I'm only wet from other causes, unrelated to umbrellas or rain. However, my umbrella causes me to be dry. If I didn't take my umbrella out, holding everything else constant, I would get wet. Here are more details and examples: https://theincidentaleconomist.com/wordpress/causation-without-correlation-is-possible/ https://theincidentaleconomist.com/wordpress/causation-witho... In this case, it's quite possible that (a) mask-wearing keeps you safe but (b) mask-wearing is only brought in when Covid is bad; and that the two countervailing effects lead to an estimated correlation of zero.
- ltbarcly3 5y ago
  Yes, if you very carefully set up the data to avoid one entire class of data then you don't have any examples of that class. When you say `mask-wearing is only brought in when Covid is bad;` that is a hypothesis that can be tested. If that is true then data rolled up at the country level would not be sufficient to determine whether masks are effective, this is a confounding effect. That is a lack of ability to find correlation in the data. The correlation is still there, your data is just biased in a way that cancels it out or has entire classes of scenarios missing. My point is that umbrellas ARE correlated with staying dry in the rain, otherwise we shouldn't bother with them as they don't work. The data collected in your example is just extremely biased. The data in this study could be biased, and I call that out. However, at a logical level, you MUST have correlation if there is causation, that is very very nearly a tautology. What does it mean to cause something if the outcome is independent?
  - dash2 5y ago
    My example wasn't set up to avoid one class of data. I said that I might get wet for other, non-rain-related reasons. So there is data with me being wet and with me being dry. And in the data, the observed correlation is zero, even though there is causation. That is because there is a confound. Also, in my toy example, the data collection process is not biased. You can collect the entire population of my days out, and you will still find no correlation. The problem is not sampling bias, it is an unobserved confound. That was just a toy example. But the linked paper may be an important real world example! You're right that "mask-wearing is only brought in when Covid is bad" is testable. Does the paper test it? Does it attempt to control for endogeneity of mask-wearing? If not, then it may be reporting a spurious null result. You're right that causation implies correlation in the following sense. "X causes Y" means that if we hold everything else equal, and change X, then Y changes. But that requires holding everything else equal. It's entirely possible that in the data we observe, we see no correlation, because an unobserved confound is biasing our estimates towards 0. That is why serious social scientists spend so much time looking for random variation, and worrying about unobserved confounds. In particular, it is just false to say "they found no correlation, so that proves no causation".