Is "impossible to say" needed for any disclaimer about trying to control for confounding factors in retrospective and observational studies?
Regarding the subject of whether vegan and vegetarian diets reduce cancer risks, I don't want to dive too deep.
But this issue seems to plague all studies of this kind, bo matter in which direction, right?
E.g. here it is an argument for doubting a positive outcome correlation (higher status, money, exercise, health attention vs diet), but it goes the other way around and is just as flawed.
E.g. stroke risk and pollution, air quality, noise, size of accomodation etc. vs income, social status and resilience.
"Impossible" to me sounds almost like a capitulation?
As if someone would say "It is impossible to say whether pollution, noise, income, diet or exercise effect stroke risk in addition to diet".
Isn't the purpose of statistics to untangle these webs of correlations.
I've found this dubious all my life, but again: in either direction, be it affirming or denying an assumed causation.
smoking and hormone supplementation are by no means independent variables when studying cancer risk. You might normalize for them by only comparing smokers with smokers or oral contraception users w other oral contraception users, but a study like that gets very complex when you have to compare smokers who exercise but use oral contraception, don't drink and are vegetarian vs smokers who exercise but use oral contraception, don't drink and are omnivores and I see no indication in the article that any of these variables were normalized for.I think you're right, the second paragraph is a warning and the word "impossible" was chosen deliberately.
Comments
They are independent variables, its their job to include the second paragraph as a warning.
Is "impossible to say" needed for any disclaimer about trying to control for confounding factors in retrospective and observational studies?
Regarding the subject of whether vegan and vegetarian diets reduce cancer risks, I don't want to dive too deep.
But this issue seems to plague all studies of this kind, bo matter in which direction, right?
E.g. here it is an argument for doubting a positive outcome correlation (higher status, money, exercise, health attention vs diet), but it goes the other way around and is just as flawed.
E.g. stroke risk and pollution, air quality, noise, size of accomodation etc. vs income, social status and resilience.
"Impossible" to me sounds almost like a capitulation?
As if someone would say "It is impossible to say whether pollution, noise, income, diet or exercise effect stroke risk in addition to diet".
Isn't the purpose of statistics to untangle these webs of correlations.
I've found this dubious all my life, but again: in either direction, be it affirming or denying an assumed causation.
smoking and hormone supplementation are by no means independent variables when studying cancer risk. You might normalize for them by only comparing smokers with smokers or oral contraception users w other oral contraception users, but a study like that gets very complex when you have to compare smokers who exercise but use oral contraception, don't drink and are vegetarian vs smokers who exercise but use oral contraception, don't drink and are omnivores and I see no indication in the article that any of these variables were normalized for.I think you're right, the second paragraph is a warning and the word "impossible" was chosen deliberately.