4 ms·

Correct me if I'm wrong... but this is just a study which finds a correlation. While the article hints at causation, does the study actually find that?

by cjensen 14y ago

streptomycin 14y ago
What would "actually find that" mean to you in this context? Science is not as clean and perfect as many would like to imagine. A strong correlation and a plausible biological mechanism combined are about as strong evidence as is possible.
- cjensen 14y ago
  People who have children later in life are going to have a number of significant demographic difference from those who have children earlier. Here's how you rule out each possibility: for each one, you pair up people in the "earlier" and "later" groups who have identical demographics. If the subgroups have identical outcomes, you have found the cause. If not, you have eliminated a cause.
  - streptomycin 14y ago
    That still doesn't "prove" causation, as there are innumerable demographics that are either impossible to match or nobody thinks to match. And FWIW, the paper does say that "factors other than father’s age do not seem to contribute substantially to the mutation rate diversity in our data".
    - learc83 14y ago
      > "factors other than father’s age do not seem to contribute substantially to the mutation rate diversity in our data". But as another poster already pointed out, age could be hiding other factors. Exposure to environmental factors that could cause mutations change with time, e.g, older fathers would have been exposed to pesticides that were banned before younger fathers were born, nutritional changes, etc...
      - streptomycin 14y ago
        I know, that's the point I was replying to in my previous post..
adgar 14y ago
It is effectively impossible to show causation in this field.
alttag 14y ago
A scientist might say, "There is evidence of causation." Typically in scientific research, to assert causation three things must be true: 1) Correlation, 2) Temporal ordering (for A to cause B, A must come before B), and 3) Sufficient theorizing, in other words, a plausible explanation, typically based on previously accepted research. In the best settings, random samples from a population with a treatment condition and a control condition are the best evidence. That's not possible in an age-base study, as randomly selected groups are not treatable with "age", and even if it were somehow possible, it probably wouldn't be ethical to do so. Unfortunately, some of the most important research can't be done with this sort of A/B test. Consequently, assertions of causation are sometimes incorrect. But that's true with any research methodology. Know that the question of causation is something the scientific community wrestles with, and a moderately informed population shouting "correlation != causation", while typically a good response to a newspaper article, isn't always the most helpful with regard to a scientific journal.

Correct me if I'm wrong... but this is just a study which finds a correlation. While the article hints at causation, does the study actually find that?

by cjensen 14y ago

Correct me if I'm wrong... but this is just a study which finds a correlation. While the article hints at causation, does the study actually find that?

streptomycin 14y ago

What would "actually find that" mean to you in this context? Science is not as clean and perfect as many would like to imagine. A strong correlation and a plausible biological mechanism combined are about as strong evidence as is possible.

cjensen 14y ago
People who have children later in life are going to have a number of significant demographic difference from those who have children earlier. Here's how you rule out each possibility: for each one, you pair up people in the "earlier" and "later" groups who have identical demographics. If the subgroups have identical outcomes, you have found the cause. If not, you have eliminated a cause.
- streptomycin 14y ago
  That still doesn't "prove" causation, as there are innumerable demographics that are either impossible to match or nobody thinks to match. And FWIW, the paper does say that "factors other than father’s age do not seem to contribute substantially to the mutation rate diversity in our data".
  - learc83 14y ago
    > "factors other than father’s age do not seem to contribute substantially to the mutation rate diversity in our data". But as another poster already pointed out, age could be hiding other factors. Exposure to environmental factors that could cause mutations change with time, e.g, older fathers would have been exposed to pesticides that were banned before younger fathers were born, nutritional changes, etc...
    - streptomycin 14y ago
      I know, that's the point I was replying to in my previous post..

adgar 14y ago

It is effectively impossible to show causation in this field.

alttag 14y ago

A scientist might say, "There is evidence of causation." Typically in scientific research, to assert causation three things must be true: 1) Correlation, 2) Temporal ordering (for A to cause B, A must come before B), and 3) Sufficient theorizing, in other words, a plausible explanation, typically based on previously accepted research. In the best settings, random samples from a population with a treatment condition and a control condition are the best evidence. That's not possible in an age-base study, as randomly selected groups are not treatable with "age", and even if it were somehow possible, it probably wouldn't be ethical to do so. Unfortunately, some of the most important research can't be done with this sort of A/B test. Consequently, assertions of causation are sometimes incorrect. But that's true with any research methodology. Know that the question of causation is something the scientific community wrestles with, and a moderately informed population shouting "correlation != causation", while typically a good response to a newspaper article, isn't always the most helpful with regard to a scientific journal.