EPTOn March 14 2012 10:43 dAPhREAk wrote:
you are criticizing the questionnaire itself, not the use of questionnaires generally (and i think your criticisms are legitimate). the methodology is always something to consider when understanding these studies.
use of "observational studies" are always important to identify issues, and where possible they should be followed up with more conclusive studies. people shouldn't just dismiss them though. the first study linking smoking with cancer was obviously an "observational study." what is this clinical study for smoking?
Show nested quote +
On March 14 2012 10:27 shinosai wrote:
Studies on smoking are different than studies done on diets. Either you smoke, or you don't. Some people smoke more than others, but it's only one thing: Smoking. But diets are multifaceted and contain a great deal more variables than smoking. It's not as simple as "either you eat red meat or you don't" because there are different kinds of red meat, and these red meats are often improperly put into a singular category. Hamburgers in the same category as pork? You bet. Grass fed and grain fed beef put in the same category? You bet.
Now if this was just a study of hot dogs, it'd be a lot simpler. Either you eat hot dogs or you don't, let's measure if there is a correlation. And of course, account for other variables that would damage our study (people who do not eat hot dogs will likely be more health conscious, not smoke, exercise, etc). But add in all these other types of food and categorize them as "red meat" and it is difficult to discern whether it's "red meat" or just particular subcategories of "red meat" causing the correlation.
Anyways, in my opinion we should never, ever use observational studies to draw conclusions in any case. I recall there was one study done where it was observed that women who took estrogen had reduced rates of cancer. As it turns out these women were more likely to exercise and were more health conscious. When a clinical study was done, taking estrogen supplements increased cancer rates by 30%! Observational studies can be used to say whatever you want. And, yes, I am saying that using observational studies to link cancer and smoking would not be sufficient evidence of causation. A clinical study would. Which we have done, and the evidence is there.
On March 14 2012 09:38 dAPhREAk wrote:
please explain to me why it was legitimate for them to use these types of survey methodologies to establish the link between cancer and smoking, but it is not proper here? because your hero is telling you not to smoke, but is poo-pooing the results of red meat studies.
On March 14 2012 09:32 spacemonkeyy wrote:
Increasing sample size actually in a lot of cases makes the study worse. If the effect is known and major i.e consuming a bottle of methylated spirits kills you all you would need would be a 1 control and a 1 variable to demonstrate the effect (obviously you would want to have a few more people to be sure of your conclusions the general accepted minimum is 50 or so). If the methodology is flawed then it doesn't matter how many people you conduct it on it is going to draw flawed conclusions- testing flawed methodology on more people doesn't undo the flaws. If your sample is too small your confidence intervals are too big to draw conclusions but if your sample is too big your confidence interval may be too small to be realistic with the accuracy of the study (i.e giving a millimeter measurement when all you can measure in is meters). Something that is not statistically significant with a small population may all of a sudden become statistically significant if you get those confidence intervals low enough with a large population. In the end you can end up picking up some extremely small effect sizes and with something as variable as diet your measuring tool (a survey every 4 years) is not accurate enough to draw conclusions of that small of a nature. For example say 1 in a million people have x outcome from consuming product A, and 2 in a million people have x outcome from consuming product B thats a 200% increase in risk. Whilst this is true statistically is it a useful or valid statement if outcome x can occur from 1000's of different confounding variables.
On March 14 2012 09:03 dAPhREAk wrote:
survey is obviously not the best research method, but it doesn't mean their conclusions are illegitimate. they used a huge sample size, which tends to validate even survey methods.
whats the other option? asking for volunteers to sit in cells for 20 years and have dieticians monitor their meat intake everyday? i'm sure people would be lining up for that study.
edit:
lol. i just read this article that you posted (http://www.marksdailyapple.com/red-meat-study/#axzz1p2fDIrR5)., which had this conclusion:
where the hell do you think we got the information on mortality rates for cigarettes from? surveys.... they didnt force people to smoke and then see what happens. although there were animal studies, for the most part researchers focus on the human surveys.
On March 14 2012 09:00 spacemonkeyy wrote:
It's one but many flaws in the study the major one being its a survey (done at every 4 years? wheres the validity in that?). Have you read the links I posted back on page 9 or 10? The evidence is all there for your perusal.
On March 14 2012 08:56 dAPhREAk wrote:
so, unless they test and account for everything possible, their study is worthless? thats not how research works. they try to limit the factors as much as possible, and consequently limit their findings to only what was tested.
On March 14 2012 08:50 spacemonkeyy wrote:
Doesn't account for orgainic vs non-organic or for what else they have in the diet plus the red meat or even how they cook the red meat. If you think of a real overweight person who is sick all the time and eats a crap diet and never moves around would you imagine his flesh is good to eat? Mass farming operations basically bring animals up in a really unnatural and sickly sort of way so that someting that would normally be rich in omega 3's (very very good for you) has barely any and is rich in omega 6 (bad). The problem is on average diet and lifestyle and the food sources are so poor in general that you need to be cautious in identifying what is actually bad about the red meat (i.e not the fact it is red).
On March 14 2012 08:43 dAPhREAk wrote:
still a 13% increase in mortality for unprocessed red meat; 20% for unprocessed red meat.
On March 14 2012 08:35 spacemonkeyy wrote:
Really this study shows what everyone deep down knows- Eat Real Food. Don't eat that processed garbage, it will literally kill you slowly. Of course processed any meat/food is bad and this study shows that the unprocessed red meat was fine. So really the issues is with the processing- although the big food companies don't want to hear that so somehow that gets lost in this all.
Really this study shows what everyone deep down knows- Eat Real Food. Don't eat that processed garbage, it will literally kill you slowly. Of course processed any meat/food is bad and this study shows that the unprocessed red meat was fine. So really the issues is with the processing- although the big food companies don't want to hear that so somehow that gets lost in this all.
After multivariate adjustment for major lifestyle and dietary risk factors, the pooled hazard ratio (HR) (95% CI) of total mortality for a 1-serving-per-day increase was 1.13 (1.07-1.20) for unprocessed red meat and 1.20 (1.15-1.24) for processed red meat.
still a 13% increase in mortality for unprocessed red meat; 20% for unprocessed red meat.
Doesn't account for orgainic vs non-organic or for what else they have in the diet plus the red meat or even how they cook the red meat. If you think of a real overweight person who is sick all the time and eats a crap diet and never moves around would you imagine his flesh is good to eat? Mass farming operations basically bring animals up in a really unnatural and sickly sort of way so that someting that would normally be rich in omega 3's (very very good for you) has barely any and is rich in omega 6 (bad). The problem is on average diet and lifestyle and the food sources are so poor in general that you need to be cautious in identifying what is actually bad about the red meat (i.e not the fact it is red).
so, unless they test and account for everything possible, their study is worthless? thats not how research works. they try to limit the factors as much as possible, and consequently limit their findings to only what was tested.
It's one but many flaws in the study the major one being its a survey (done at every 4 years? wheres the validity in that?). Have you read the links I posted back on page 9 or 10? The evidence is all there for your perusal.
survey is obviously not the best research method, but it doesn't mean their conclusions are illegitimate. they used a huge sample size, which tends to validate even survey methods.
whats the other option? asking for volunteers to sit in cells for 20 years and have dieticians monitor their meat intake everyday? i'm sure people would be lining up for that study.
edit:
lol. i just read this article that you posted (http://www.marksdailyapple.com/red-meat-study/#axzz1p2fDIrR5)., which had this conclusion:
The real take home message from this study is this: Don’t be obese, do exercise, don’t smoke, eat plenty of vegetables and fruit, take supplements, avoid processed meats, avoid overcooked meats, eat from a variety of animal foods.
where the hell do you think we got the information on mortality rates for cigarettes from? surveys.... they didnt force people to smoke and then see what happens. although there were animal studies, for the most part researchers focus on the human surveys.
Increasing sample size actually in a lot of cases makes the study worse. If the effect is known and major i.e consuming a bottle of methylated spirits kills you all you would need would be a 1 control and a 1 variable to demonstrate the effect (obviously you would want to have a few more people to be sure of your conclusions the general accepted minimum is 50 or so). If the methodology is flawed then it doesn't matter how many people you conduct it on it is going to draw flawed conclusions- testing flawed methodology on more people doesn't undo the flaws. If your sample is too small your confidence intervals are too big to draw conclusions but if your sample is too big your confidence interval may be too small to be realistic with the accuracy of the study (i.e giving a millimeter measurement when all you can measure in is meters). Something that is not statistically significant with a small population may all of a sudden become statistically significant if you get those confidence intervals low enough with a large population. In the end you can end up picking up some extremely small effect sizes and with something as variable as diet your measuring tool (a survey every 4 years) is not accurate enough to draw conclusions of that small of a nature. For example say 1 in a million people have x outcome from consuming product A, and 2 in a million people have x outcome from consuming product B thats a 200% increase in risk. Whilst this is true statistically is it a useful or valid statement if outcome x can occur from 1000's of different confounding variables.
please explain to me why it was legitimate for them to use these types of survey methodologies to establish the link between cancer and smoking, but it is not proper here? because your hero is telling you not to smoke, but is poo-pooing the results of red meat studies.
Studies on smoking are different than studies done on diets. Either you smoke, or you don't. Some people smoke more than others, but it's only one thing: Smoking. But diets are multifaceted and contain a great deal more variables than smoking. It's not as simple as "either you eat red meat or you don't" because there are different kinds of red meat, and these red meats are often improperly put into a singular category. Hamburgers in the same category as pork? You bet. Grass fed and grain fed beef put in the same category? You bet.
Now if this was just a study of hot dogs, it'd be a lot simpler. Either you eat hot dogs or you don't, let's measure if there is a correlation. And of course, account for other variables that would damage our study (people who do not eat hot dogs will likely be more health conscious, not smoke, exercise, etc). But add in all these other types of food and categorize them as "red meat" and it is difficult to discern whether it's "red meat" or just particular subcategories of "red meat" causing the correlation.
Anyways, in my opinion we should never, ever use observational studies to draw conclusions in any case. I recall there was one study done where it was observed that women who took estrogen had reduced rates of cancer. As it turns out these women were more likely to exercise and were more health conscious. When a clinical study was done, taking estrogen supplements increased cancer rates by 30%! Observational studies can be used to say whatever you want. And, yes, I am saying that using observational studies to link cancer and smoking would not be sufficient evidence of causation. A clinical study would. Which we have done, and the evidence is there.
you are criticizing the questionnaire itself, not the use of questionnaires generally (and i think your criticisms are legitimate). the methodology is always something to consider when understanding these studies.
use of "observational studies" are always important to identify issues, and where possible they should be followed up with more conclusive studies. people shouldn't just dismiss them though. the first study linking smoking with cancer was obviously an "observational study." what is this clinical study for smoking?
Well of course all questionaires aren't "BAD" each one had to be evaluated on it's own merits. A lot of areas of research use questionaires (i.e to measure pain and disability) but they have to be validated first before they are used. If the questionaire hasn't been validated or standardised and doesn't seem particualrly accurate its its description then I'm sure you can forgive us for thinking that this type of survery is low quality.
