Chris,
Any poll needs to start or end with who is answering the questions i.e. fitness swimmer, triathlete, occasional competitor, compulsive competitor etc.
I am not sure of the statistics on this, but for any poll to be statistically valid/significant, it would need to have what, 60% of the given population? If that is true, than any poll we do here is not likely to rise to that threshold if we are talking either about the entire USMS population or any subgroup of the population. What percentage of those in USMS follow the forum?
Surely there is someone on the forum who knows enough about statistics to give us some insight on this. What would be the number needed for a representative sample? Would there need to be a representative sample for each of the subgroups mentioned above?
As Julie says, a sample can be representative without being large. How representative it is depends much more on how the sample is chosen, rather than the size. The forum as a sample is biased not because there are relatively few of us, but because among other things we tend to be a little more passionate (okay, weird) about swimming than a typical USMS member. But a relatively large sample would still not be representative if it was chosen poorly.
A large representative sample is better -- but more expensive -- than a small representative sample only in that it improves precision, decreasing those "margins of error" you always hear about in polls.
As far as the necessity of a poll asking for the data you mention, that depends entirely on the purpose of the poll. "Throwing out a question" is fine if you want to get a sense for how the majority feels about that question, regardless of how "competitive" that majority is.
The only reason I can see to ask for the information you want is to see how it breaks out by proficiency and desire to compete. Why do you want to know that? I am a little wary of asking for that information if it will (for example) cause non-competitors to avoid taking the poll for fear their voice will not "count." (Which would cause the sampling bias you seek to avoid.) Possibly I could include such questions at the end of the poll and mark it as optional.
Electronic-only polls might introduce a sampling bias too.
Chris,
Any poll needs to start or end with who is answering the questions i.e. fitness swimmer, triathlete, occasional competitor, compulsive competitor etc.
I am not sure of the statistics on this, but for any poll to be statistically valid/significant, it would need to have what, 60% of the given population? If that is true, than any poll we do here is not likely to rise to that threshold if we are talking either about the entire USMS population or any subgroup of the population. What percentage of those in USMS follow the forum?
Surely there is someone on the forum who knows enough about statistics to give us some insight on this. What would be the number needed for a representative sample? Would there need to be a representative sample for each of the subgroups mentioned above?
As Julie says, a sample can be representative without being large. How representative it is depends much more on how the sample is chosen, rather than the size. The forum as a sample is biased not because there are relatively few of us, but because among other things we tend to be a little more passionate (okay, weird) about swimming than a typical USMS member. But a relatively large sample would still not be representative if it was chosen poorly.
A large representative sample is better -- but more expensive -- than a small representative sample only in that it improves precision, decreasing those "margins of error" you always hear about in polls.
As far as the necessity of a poll asking for the data you mention, that depends entirely on the purpose of the poll. "Throwing out a question" is fine if you want to get a sense for how the majority feels about that question, regardless of how "competitive" that majority is.
The only reason I can see to ask for the information you want is to see how it breaks out by proficiency and desire to compete. Why do you want to know that? I am a little wary of asking for that information if it will (for example) cause non-competitors to avoid taking the poll for fear their voice will not "count." (Which would cause the sampling bias you seek to avoid.) Possibly I could include such questions at the end of the poll and mark it as optional.
Electronic-only polls might introduce a sampling bias too.
