Top
- The sampling investigation is not investigation in an all-round way, it selects some samples from the research object and then to make estimation or diagnosis. Obviously sampling investigation choose units to reflect overall information, therefore, it can play a role in investigating in an all-round way.
- According to methods to select samples, sampling investigation can be divided into probability sampling and none probability sampling. Probability sampling is conducted on the basis of probability theory and mathematical statistics.It selects samples according to random principles, and it infers to estimate to make some characteristics in quantity. In our country we are accustomed to name them sampling investigation
- There are three outstanding characteristics in the sampling investigation:
- 1, Select samples according to the random principles;
- 2, Every unit has certain probability to be chosen in overall
- 3,Can guarantee to control the error within the fixed range with certain probability.
- Several kinds of concrete sample ways
1, Simple random sampling : It is also called simplicity random sampling too, which means to collect n units from overall N randomly and each possible sample release to be an equal opportunity to be chosen. According to the frequency to be selected this method includes not repeated sampling and repeated sampling. In the sampling investigation especially in the sampling investigation of social economy, the simple random sampling generally means not repeated sampling. The simple random sampling is the foundations of other sample methods, because it is easiest to deal with in theory. But in reality, when N is quite big, the simple random sampling is not very easy to do. It demands a frame that includes all N samples at first; Samples comparatively are scattered secondly and thirdly investigation is not easy to implement. So, it is actually few to be adopted directly in reality.
2, Layered Sampling: It is also called classified sampling or type sampling. It first divides overall group into k parts and all of which have none in over 2 parts, and the part is called layer. Then select n1 , n2......,nk separately in each layer.
The functions of dividing layer are three aspects: First, for the convenience and research intention of the purpose of the work; Second, in order to improve the precision of sampling; Third, for under certain request of precision, reduce unit of sample count by economizing and investigating the expenses. So dividing layer is a very general sampling technology.
In fact, layered sampling combines scientific grouping and sampling theory, the former divides layers where nature is relatively close in order to reduce the variation degree between the marking value; The latter selects samples according to the principle of sampling. So, layered sampling can get more accurate inference results through surveying less samples. When overall figures larger, more complicated inside structure, layered sampling can make the satisfactory result.
3, Cluster sampling: It first divides the group into several non-cross, non-repeated group and we call them clusters. Then the group will be regarded as the sampling units. Cluster sampling is mainly used in framework that is absent from overall units. When to use cluster sampling, the group should possess a better representative, namely heavy group differences but small differences among groups. The advantages of this method is that it is convenient, saving the funds. And Its disadvantage is that it has a higher deviation than other methods.
Stratified systematic sampling:
A form of probabilistic sampling which combines elements of (1) simple random sampling, (2) stratified random sampling, and (3) systematic sampling, in an effort to reduce sampling bias.
Stratified sampling is commonly used probability method that is superior to random sampling because it reduces sampling error. A stratum is a subset of the population that share at least one common characteristic. Examples of stratums might be males and females, or managers and non-managers. The researcher first identifies the relevant stratums and their actual representation in the population. Random sampling is then used to select a sufficient number of subjects from each stratum. "Sufficient" refers to a sample size large enough for us to be reasonably confident that the stratum represents the population. Stratified sampling is often used when one or more of the stratums in the population have a low incidence relative to the other stratums.
5. Multi-stage sampling
One of the problems of simple random sampling is that if the sample selected is widely scattered over the country the interviewer may spend more time in traveling than he does in actually interviewing. Often, then, the cost of taking such a sample can be prohibitive. This is especially true in countries in which urbanization has not yet developed. You can imagine the problems involved if you were to take a sample of people scattered all over central Canada, Malaysia or West Africa. Interviewers might have to spend weeks in traveling. It is to overcome this problem that multi-stage sampling has been developed.
Rather than spending time in traveling to interview 2000 people scattered all over Malaysia, is it more convenient to interview say 100 people in each of twenty selected areas. Provided that we can ensure that the sample ultimately selected is representative, a great deal of time and expense can be saved.
In many ways the selection of a multi-stage sample is similar to the selection of a stratified sample, but we select primary groups and sub-groups geographically rather than on the basis of social characteristics. Typically, the first stage is to break down the area under survey into a number of standard regions. These may be areas such as the English county, the Canadian province or any other easily defined administrative area. The sample is then divided among these regions according to their population. The second stage is to select at random a small number of districts, say towns and villages, within the primary region. Once again it is necessary to allocate to each of these districts a number of interviews proportional to its population. Almost certainly at this stage, an element of stratification must appear if the sample is to be representative.
Let us suppose that we are selecting six towns at random within each region. If one third of the population live in cities of over 200,000 people, a further third in towns with a population of from 20,000 to 200,000 and the remainder in small towns and villages of under 20,000 inhabitants, it would be desirable to select two areas of each type. Finally, having selected our towns and villages within each region, the sample in each town or village is chosen by some random method. Again, it may be felt to be desirable to choose a stratified sample within each town. How far this stratification is carried depends, of course, on the purpose of the survey and homogeneity of the population. If we are examining housing conditions, and the type of housing ranges from one-roomed flats to 25-roomed mansions, some degree of stratification will be necessary. But if all the housing in the area is of an essentially similar type, as it may be in a small village, stratification is probably not necessary.
6, Double sampling is the single best way to estimate volume. It takes advantage of the characteristics of point sampling to dramatically speed up a cruise without sacrificing accuracy. Double sampling calculations are actually easier than the usual point sampling calculations since information requirements are reduced.
7, Quota Sampling:
There can be few people who have not been stopped in the street by an interviewer holding a questionnaire, or who at least know someone who has been so stopped. And the immediate reaction seems to be, `Why did they chose me?' The essence of this type of sample is that it is not preselected, but is chosen by the interviewer on the spot. He, or more usually she, is given a certain number (a quota) of questionnaires which it is her job to have completed during the course of the week, or month. She has a free choice of whom she asks to answer the questions. This choice, naturally, is subject to some general restrictions. One can imagine a young male interviewer stopping only girls of about his own age whom he thought to be pretty. Free choice does not extend as far of this. The interviewer is told that the completed quota of questionnaires must include a certain number of males and of females; a certain number from specified age groups such as, under 18, 18 to 40, 40 to 65. Controls such as this are easy to implement, but often the quota has to be further subdivided by social class or occupation. This can make quota sampling very difficult as there is no real definition of what constitutes, say, the upper middle class, and certainly you can seldom tell a person's social class by looking at him. Such a requirement necessitates the interviewer being given detailed definitions and descriptions of what the survey body means by each term it uses.
Subject, to the above controls, the sample is chosen by the interviewer from people who pass in the street. Now, obviously, such a technique of sampling can be open to a great deal of abuse and bias. When it was first introduced, many cases were reported of interviewers sitting at home filling in their own questionnaires without ever interviewing anyone. Of course this was soon remedied by the questionnaire asking for names and addresses, and the survey body making a spot check to ensure that such people had actually been interviewed. Deliberate fraud of this nature is very rare now.
Much more important is the bias to which the quota survey may be subject. If the interviewer does not start work until, say, ten o'clock, the majority of people she can stop must of necessity be the housewife or the unemployed; if she starts at, say, nine o'clock, the vast majority will be clerical or managerial workers; to start even earlier means that she will probably meet largely manual workers. Thus, the interviewer must be prepared to spread her work throughout the day. It is little use trying to have all the questionnaires completed before the morning coffee-break. A further source of bias is that the people you meet in the street are either going somewhere or doing something; they resent being interrupted and their answers may be hurried and slapdash, given without serious thought. These problems necessitate further controls and advice being given to the interviewer. She may be told at what times to conduct the interviews or even where to conduct them. All this is not to imply that interviewers are rascals or rogues - far from it - but it must be pointed out that the success of such a survey depends very much on their following their instructions to the letter.
Since the quota sample has these weaknesses, why has it become increasingly common? Well, for one thing, it is cheap. It eliminates repeated calls to interview a person who may not be in at home on the first two or three occasions. In fact, it is estimated that each interview in a quota sample costs only about half as much as each interview in a random sample. Then, too, we have in every survey a number of people who do not respond. In a quota sample this may lead to serious bias. With quota sampling a person not willing to co-operate is ignored at once and we pass straight on to the next person with no loss of time and no cost. In other types of sampling a substitute may have to be found for each person who does not respond. It might, in fact, be argued that this itself introduces bias into the sample. It may be that the man who is prepared to fill in a questionnaire is some kind of extrovert, and we may be getting the views of only one type of individual. Nevertheless, given the existence of the controls and checks that have been mentioned, all the evidence points to the fact that in skilled hands quota sampling gives reasonably satisfactory results.
Statisticians generally accept that this type of quota sampling is not a substitute for random sampling statistically, but it is so quick and so cheap a method of carrying out surveys that it is not likely to be replaced for a very long time.
All the methods lined up should be applied to the practice, greatly adopting experiences and theories. And our intention is to reduce error and make data and information closer to the fact.
|