Sampling

The first stage in sampling is to choose the target population that you intend to investigate.  It would be impossible to survey the entire student population of LCC for your dissertation/final project so you need to consider how to get a representative view from a smaller number of students. To do this you need to consider whether you take a probability sample or a non-probability sample.

There is no set rule about how many cases you need to include but a good rule of thumb is 30 (Economist, 1993) for each category as this provides you with a large enough sample to be able to break it down into smaller sub-samples if required, and also to perform statistical analysis. 

Probability sampling requires you to use a sampling frame from which you can draw a random sample for your research. We could work with a sampling frame in the form of the full listing of students attending LCC.  Remember, however, that if the list is incomplete then the sample you draw may not be representative of the whole population.  The sample size needs to be large enough to provide data which you can analyse and draw generalisations from, obviously the larger it is the more accurate your data.  As you have neither the time nor the resources to investigate every student we would use a random quota.
You need to collect enough data which can be analysed using the methods you have chosen and which will produce meaningful results.  As we have several different Schools within the college if you are wanting to generalise the findings to the whole College you would have to ensure that you have sampled from every School.  To do this the most common approach would be to take a Simple Random Sample from each School list, perhaps making the profile more accurate by proportionate sampling so that the bigger the School the more students you survey.

A Stratified Sample would be drawn if you decided to group the students by School and course and then draw a random sample from each.  By doing this you will be able to make statements which are more representative of those strata but not necessarily of the whole population.
 
A Cluster Sample is slightly different in that we could choose to sample students at the E&C site who are on three particular courses.  Data would be collected from all the students on those courses but this data would be less representative of the whole population than that collected using the previous methods.

A quota is the usual method of non-probability sampling used in
research and is particularly useful in the pilot stage to test the questions etc…  Having collected the student lists you would decide to sample a given number from each course/year and set out to interview that number of students.  This is the most common sampling technique used in market research where companies want to gather information about certain consumer groups.  The researchers interview individuals who fit the criteria of their brief, every wondered why you are not stopped in the street when everyone else seems to be?#p#分页标题#e#

A purposive sample is drawn when you use your judgement to decide who you will ask to participate in the research – if you are looking to get information about the use students make of the library you might decide to speak to one group of students you often see there, and another group who you know go there less often.  This way you will be getting data from two ends of the continuum of library use.

We often find ourselves working with a self-selecting sample, such as your friends and colleagues who offer to be interviewed for your research because they have a view they would like to express.
Closer to home we may even have a convenience sample if we choose the students in our class because we have access to them. 
At the pilot stage a non-probability sample is acceptable but you need to refer to the fact that findings from such a sample will be less representative than those from a probability sample.

本文由www.ukthesis.org整理