4 ms·

Why don't you read the study and raise the specific confounders you think are missing? The answer is universally "no", there's no such thing as "controlling fo

Why don't you read the study and raise the specific confounders you think are missing? The answer is universally "no", there's no such thing as "controlling for [all] confounding variables" without blinded randomization. But we unfortunately have to rely on observational studies to try to understand the world. At least to understand what specific questions can justify the cost and complexity of a randomized study. So yeah, of course there are confounders. Some are controlled in this study and others are not. Which are the important ones you think are missing?

bcjdjsndon 3mo ago
[flagged]
- estearum 3mo ago
  What gave you the impression I'm mad? I'm providing a useful comment explaining why the answer to the zero-value "controlling for confounders right? harharhar" top Reddit comment is always "no." Unless it's a randomized study, the answer is no. But that doesn't mean it doesn't tell us information. This is helpful information for anyone who wants to transcend midwit internet snark and understand how scientific knowledge is built.
  - qwerpy 3mo ago
    I appreciated your comment and especially liked the way you characterized the original commenter. Everyone going for low effort points here is cheapening the discussion, like that R website we hate being compared to.
    - bcjdjsndon 3mo ago
      At least there's the occasional cat on reddit
  - bcjdjsndon 3mo ago
    > midwit I'll take that as a compliment
    - estearum 3mo ago
      You should! Indeed it was generous.
- vivekd 3mo ago
  I think he had a reasonable point, I don't know if he was mad or not but that doesn't take away from the validity of his statement. Simply saying a confounding variable was missing isn't helpful because it's impossible to control for all confounding variables. If however there was an important variable someone can think of, like maybe parents who control their kids phone use could also be more involved in their children's learning that could be a confounding variable worth mentioning to think about how valid the study is - but even then I don't know if that by itself would be enough to dismiss the study
  - bcjdjsndon 3mo ago
    > Simply saying a confounding variable was missing isn't helpful because it's impossible to control for all confounding variables That's the problem with studies though. My point is why defend something so ardently when we all agree it's shit but the best we have?
    - estearum 3mo ago
      Uhh... there's a truly gigantic gap between the shit end of "has confounders" and the good end of "has confounders", and locating where a particular study lands on that spectrum is a foundational skill to evaluating virtually any factual claim you'll ever encounter... I am advocating for people to identify that there's a spectrum instead of just saying "oh well, every study is imperfect" which diverts attention from understanding in what particular ways is it imperfect.
      - bcjdjsndon 3mo ago
        > is a foundational skill to evaluating virtually any factual claim you'll ever encounter... Calm down Doris. You advocate for what you like but I'm just not convinced
rahimnathwani 3mo ago
Sure, a drive-by "did they control for confounders???" without naming any is low-effort, and it's reasonable to ask GP to be specific. BUT you overshot in the other direction! "There's no such thing as controlling for confounders without blinded randomization" isn't fair. Sure, randomization is the cleanest identification strategy, but it's not the only one. Instrumental variables, regression discontinuity, difference-in-differences, sibling or twin fixed effects, and natural experiments are all used in economics, and they exist because randomization is often impossible or unethical, and you have to rely on natural experiments. (Separately, a phone-ownership study of 9yo kids has an obvious potential for selection bias. So in a sense I'm sympathetic to GP asking about confounders without naming any.
- estearum 3mo ago
  If you actually copy/pasted the "quoted" text from my comment, you will have found I said something different from what you claimed. Virtually any study has an obvious potential for selection bias. That's why people write up how they did the study and then other people can interrogate and criticize it. Might as well just paste a link to https://en.wikipedia.org/wiki/Bias_(statistics)#Types https://en.wikipedia.org/wiki/Bias_(statistics)#Types in every comment section ever.

bcjdjsndon 3mo ago

[flagged]

estearum 3mo ago
What gave you the impression I'm mad? I'm providing a useful comment explaining why the answer to the zero-value "controlling for confounders right? harharhar" top Reddit comment is always "no." Unless it's a randomized study, the answer is no. But that doesn't mean it doesn't tell us information. This is helpful information for anyone who wants to transcend midwit internet snark and understand how scientific knowledge is built.
- qwerpy 3mo ago
  I appreciated your comment and especially liked the way you characterized the original commenter. Everyone going for low effort points here is cheapening the discussion, like that R website we hate being compared to.
  - bcjdjsndon 3mo ago
    At least there's the occasional cat on reddit
- bcjdjsndon 3mo ago
  > midwit I'll take that as a compliment
  - estearum 3mo ago
    You should! Indeed it was generous.
vivekd 3mo ago
I think he had a reasonable point, I don't know if he was mad or not but that doesn't take away from the validity of his statement. Simply saying a confounding variable was missing isn't helpful because it's impossible to control for all confounding variables. If however there was an important variable someone can think of, like maybe parents who control their kids phone use could also be more involved in their children's learning that could be a confounding variable worth mentioning to think about how valid the study is - but even then I don't know if that by itself would be enough to dismiss the study
- bcjdjsndon 3mo ago
  > Simply saying a confounding variable was missing isn't helpful because it's impossible to control for all confounding variables That's the problem with studies though. My point is why defend something so ardently when we all agree it's shit but the best we have?
  - estearum 3mo ago
    Uhh... there's a truly gigantic gap between the shit end of "has confounders" and the good end of "has confounders", and locating where a particular study lands on that spectrum is a foundational skill to evaluating virtually any factual claim you'll ever encounter... I am advocating for people to identify that there's a spectrum instead of just saying "oh well, every study is imperfect" which diverts attention from understanding in what particular ways is it imperfect.
    - bcjdjsndon 3mo ago
      > is a foundational skill to evaluating virtually any factual claim you'll ever encounter... Calm down Doris. You advocate for what you like but I'm just not convinced

rahimnathwani 3mo ago

Sure, a drive-by "did they control for confounders???" without naming any is low-effort, and it's reasonable to ask GP to be specific. BUT you overshot in the other direction! "There's no such thing as controlling for confounders without blinded randomization" isn't fair. Sure, randomization is the cleanest identification strategy, but it's not the only one. Instrumental variables, regression discontinuity, difference-in-differences, sibling or twin fixed effects, and natural experiments are all used in economics, and they exist because randomization is often impossible or unethical, and you have to rely on natural experiments. (Separately, a phone-ownership study of 9yo kids has an obvious potential for selection bias. So in a sense I'm sympathetic to GP asking about confounders without naming any.

estearum 3mo ago
If you actually copy/pasted the "quoted" text from my comment, you will have found I said something different from what you claimed. Virtually any study has an obvious potential for selection bias. That's why people write up how they did the study and then other people can interrogate and criticize it. Might as well just paste a link to https://en.wikipedia.org/wiki/Bias_(statistics)#Types https://en.wikipedia.org/wiki/Bias_(statistics)#Types in every comment section ever.