4 ms·

It's a fascinating article. Even more fascinating: nobody here in the comments on the most skeptical site I know has pointed out that the article fails to prov

It's a fascinating article. Even more fascinating: nobody here in the comments on the most skeptical site I know has pointed out that the article fails to provide even a hint of the degree to which it alters those perceptions. Is it 1%? 10%? 30%? 80%? Kind of an important detail.

ordu 6y ago
How to measure the degree? You've wrote percentages as an example, how they should be understood? The frequency of altered behavior? But there are tons of different factors altering behavior, as personal so situational factors. They somehow also should be measured. I would be much more skeptical, if they provided measurements outside of ordinal scale of measurement[1]. One can compare two setups on basis less/greater, but not to measure them in absolute scale. [1] https://en.wikipedia.org/wiki/Level_of_measurement https://en.wikipedia.org/wiki/Level_of_measurement
- aidenn0 6y ago
  Using standard-deviations would be good, assuming risk-assessment is a normal distribution.
  - ordu 6y ago
    It would be standard deviations in some specific situation. Maybe in 2-3 situations. Like answering questions of a psychologist and inflating a balloon in a kind of a computer game. Does it influence a driving style of a person? People could overcome biases, and possibly driving as a habit would be more strong than acetaminophen. The only reason to go numeric with this study is to measure statistical significance of their findings. I'm sure they did it, but their numbers have no meaning outside of their study.
    - aidenn0 6y ago
      Numeric is also a good sanity check. If they claim e.g. 3 standard deviations of risk-taking increase on typical doses of paracetamol, then there's something wrong with the study, because we would have noticed that already.
      - ordu 6y ago
        I'm sure they did all the math. It is just journalists do not report it, because it's too boring. And in this case this math would be meaningless if you didn't dig into methods, experimental plan, number of participants, how they where split into groups, and so on. With this numbers one would need to report all the science article. If you interested in such details, you should go to the original article. Speaking of skepticism, I personally just believe, that article is good enough. Mostly they are. In this case I see no methodological troubles, the experiment seems straightforward, I would have no troubles designing it, based on this journalist's article and I'm not a professional scientist (though I have a psychology degree). > If they claim e.g. 3 standard deviations of risk-taking increase on typical doses of paracetamol, then there's something wrong with the study, because we would have noticed that already. If they claimed 3 std-devs of increase, it would be 3 std-devs of increase in a specific model situation of risk taking from the experiment. It would be great news, it would mean that researchers found a good model situation with high sensitivity to a paracetamol. Would be nice to see how some other factors influencing risk taking show themselves in this setup.
taneq 6y ago
I've been reading a bunch of medical studies recently and I've found the same thing painfully often. Wow, this study is p<0.01, that's amazing! How much did the substance alter the measurement, though? Never mentioned.

ordu 6y ago

How to measure the degree? You've wrote percentages as an example, how they should be understood? The frequency of altered behavior? But there are tons of different factors altering behavior, as personal so situational factors. They somehow also should be measured. I would be much more skeptical, if they provided measurements outside of ordinal scale of measurement[1]. One can compare two setups on basis less/greater, but not to measure them in absolute scale. [1] https://en.wikipedia.org/wiki/Level_of_measurement https://en.wikipedia.org/wiki/Level_of_measurement

aidenn0 6y ago
Using standard-deviations would be good, assuming risk-assessment is a normal distribution.
- ordu 6y ago
  It would be standard deviations in some specific situation. Maybe in 2-3 situations. Like answering questions of a psychologist and inflating a balloon in a kind of a computer game. Does it influence a driving style of a person? People could overcome biases, and possibly driving as a habit would be more strong than acetaminophen. The only reason to go numeric with this study is to measure statistical significance of their findings. I'm sure they did it, but their numbers have no meaning outside of their study.
  - aidenn0 6y ago
    Numeric is also a good sanity check. If they claim e.g. 3 standard deviations of risk-taking increase on typical doses of paracetamol, then there's something wrong with the study, because we would have noticed that already.
    - ordu 6y ago
      I'm sure they did all the math. It is just journalists do not report it, because it's too boring. And in this case this math would be meaningless if you didn't dig into methods, experimental plan, number of participants, how they where split into groups, and so on. With this numbers one would need to report all the science article. If you interested in such details, you should go to the original article. Speaking of skepticism, I personally just believe, that article is good enough. Mostly they are. In this case I see no methodological troubles, the experiment seems straightforward, I would have no troubles designing it, based on this journalist's article and I'm not a professional scientist (though I have a psychology degree). > If they claim e.g. 3 standard deviations of risk-taking increase on typical doses of paracetamol, then there's something wrong with the study, because we would have noticed that already. If they claimed 3 std-devs of increase, it would be 3 std-devs of increase in a specific model situation of risk taking from the experiment. It would be great news, it would mean that researchers found a good model situation with high sensitivity to a paracetamol. Would be nice to see how some other factors influencing risk taking show themselves in this setup.

taneq 6y ago

I've been reading a bunch of medical studies recently and I've found the same thing painfully often. Wow, this study is p<0.01, that's amazing! How much did the substance alter the measurement, though? Never mentioned.