Representative sample and science
Delighted to see a debate published in August 2013 issue of the Int J Epidemiol (IJE) on how representative sample from a defined population helps or scuttles scientific research. I am quoting from an article by Rothman and others:
"Here we suggest that representativeness may be essential for conducting opinion polls, or for public-health applications, but it is not a reasonable aim for a scientific study."
"Clearly, representativeness does not, in and of itself, deliver valid scientific inference. If a study population is representative of some larger source population, the overall associations observed in the study population may not apply to every subgroup. The overall effect is merely an average effect that has been weighted by the distribution of people across these subgroups. Thus, if you have a sample that is representative of the sex distribution in the source population, the results do not necessarily apply either to males or to females, but only to a hypothetical person of average sex."
"Surveys vs causal studies: One way to distinguish science from the kind of information that surveys produce is its overall applicability in space and time. Scientific statements ideally serve to describe nature in a way that is not limited to one time and one place"
Those subscribing to IJE can log on to see http://ije.oxfordjournals.org/content/42/4/1012.full
for details.
Delighted to see a debate published in August 2013 issue of the Int J Epidemiol (IJE) on how representative sample from a defined population helps or scuttles scientific research. I am quoting from an article by Rothman and others:
"Here we suggest that representativeness may be essential for conducting opinion polls, or for public-health applications, but it is not a reasonable aim for a scientific study."
"Clearly, representativeness does not, in and of itself, deliver valid scientific inference. If a study population is representative of some larger source population, the overall associations observed in the study population may not apply to every subgroup. The overall effect is merely an average effect that has been weighted by the distribution of people across these subgroups. Thus, if you have a sample that is representative of the sex distribution in the source population, the results do not necessarily apply either to males or to females, but only to a hypothetical person of average sex."
"Surveys vs causal studies: One way to distinguish science from the kind of information that surveys produce is its overall applicability in space and time. Scientific statements ideally serve to describe nature in a way that is not limited to one time and one place"
Those subscribing to IJE can log on to see http://ije.oxfordjournals.org/content/42/4/1012.full
for details.