The question cannot be answered without first determining WHICH population is to be sampled reliably? Is "population" intended to mean "the population of all cinema-going people?" On this interpretation, the question asks, "how often do those who go to cinema typically go (per week, month, or other unit of time as "rate" implies)?' On this interpretation, surveying outside the cinema is by far the best method, because it selectively questions only members of the cinema-going population. It is questionable whether a "rate" based on any other population would be meaningful. If we discovered that the total number of movie visits in, say, Burma, represents a population rate of 0.023 times per week, this tells us nothing of how often Burmese people go to the cinema. Chances are, vast numbers of people almost never go, while some small group of people go with some regularity. Only the latter are really relevant, unless its a question about market saturation, in which case rate is the wrong measure.....I would think you'd first want a head count (e.g., movie-goers per capita).
A postal survey is a loser because reliability depends upon the representativeness of the returns. People who return postal surveys constitute a self-defined sub-population of the specified group, and return rates tend to be dismal, especially on such trivial kinds of impersonal data. I'm not beaurocratically sophisticated enough to know exactly what is meant by a register, but I take it to be a town-hall listing of residence occupancy that would likely be fairly accurate in identifying who lives where. I think I might regard the census or registry in principle as the best definition of the general (geographic/national/regional) population for purposes of sampling. The census is compiled for the purpose of listing everyone (it strives to be fully inclusive) which, though imperfect, is by definition what is required for a proper random sampling (identify all possible members of the population and ensure each has an equal probability of being selected). That it is a "household survey" says we go to them, and are not dependent upon the number (and type) of people who choose to return their responses. I think 3 is therefore the best answer if the term "population" is intended to mean "geopolitical body," and the concerns mentioned above can be dismissed (viz a viz, the cinema-going population). A house to house survey in the evening would not be better than a sampling of people listed in a census or registry because it further requires that people be home in the evening. People who go out in the evening (for example, cinema-goers) would be unable to answer the survey, so a hopeless bias would be introduced. I would be tempted to administer a question like this in essay form precisely to see if the student understands the first critical issue--defining the population of interest. I'd say #4 if population is geopolitically defined, or #3 if we are specifically interested only in the cinema-going population. I hope you will post the anwer. Dessie !!!!!!!!!!!!!!!STAT IS LIFE!!!!!!!!!!!!!!!!!!! On Tue, 12 Mar 2002 15:14:28 +0100, George <[EMAIL PROTECTED]> wrote: >from a past exam, I have the following question: > >Which type of sample survey would provide the most reliable data on the >frequency of cinema-going of the population? > >only one of 1,2,3,4 is "correct": > >1) A house to house survey in the evening >2) A postal survey based on the electoral list >3) Interviewing outside the cinema >4) A household survey based on a population census or register > >------- >I can assume cost of the survey is not an issue right? then 2 is out, as >1 is better than 2 for reliability. >3 is out of the question as it does not cover the general population. > >So I think either 1 or 4 must be correct. >But 4 is similar to 1, so what is the difference between a census and a >Register??? and between a Register and a (i.e. electoral) List??? >BTW a household survey is also presumably done from house to house, and >not via post ? > >Thanks for your help. particularly, Dr. Paul Gardner for his past help >in a similar question >on finite sampling. > >George > >-for personal reply please remove SPAMLESS from my email address. > . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
