Sampling in quantitative research

Figure 8.9 displays the “four sources of error in social survey research” (p194), including ‘data-processing’ error. As the term implies, this is an error which occurs at the time of processing the data rather than at the time of preparing for it or even gathering it. The typical processing error crops up in coding answers given in questionnaires. It is true that faulty questionnaire construction may ‘breed’ errors at the processing stage, so that great care must be taken at the implementation phase and while there is, indeed, a standard ‘error’ between the averages of samples and populations, this is a statistical expression rather than a human mistake.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 194

The standard error is that which can be calculated as the difference between the population average and the sample average. Once the sample has been selected randomly, we can determine the probable difference between the sample and the population as a whole, as a range. We usually express our results, therefore, with a high degree of confidence (but not total) that our results apply to the entire population, plus or minus a little. It sounds more tentative than we might like but it cannot be more accurate than that. It should be pointed out that stratification of a sample can reduce the standard error.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 182,183

Sampling theory (see fig 8.8 on p182) tells us that sampling error is measured in terms of the ‘standard error of the mean’, which means, briefly, that there will always be a high probability of having a sampling error of a particular size. By comparing the standard error in our own research (in other words, the standard deviation in our own sample from the simple average) with the generally expected standard error, we can arrive at the actual sampling error of our own research. This may sound complicated but, like question 4, our concern should be with claiming for our research findings only what can be fairly and honestly applied to the entire population. We can increase the size of our sample to reduce the sampling error but, unless we research the entire population, we can never eliminate it. This is actually good news for researchers because a sample can actually be quite small and still yield good results, “plus or minus a certain %”.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 182-184

A frame is a surround for something, like a frame for a photograph or a university degree, which we hang on our walls. A sampling frame ‘surrounds’ the population we want to study in our research. We won’t usually have the time or the money to ask questions of each member of the population, so we will interview or survey only a limited number of people. How do we know that the people we interview are truly representative of the entire population? Usually we don’t know for sure but we have a better chance if we select people at random from particular sections of the population, so that we can, at least, say our sample represents all sections of the population as they showed up in our overall ‘picture’, our sampling frame.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 174

Sometimes it is very difficult to produce a sampling frame for the population we wish to study, in which case probability sampling is not easily available to us. Since this automatically impairs generalizability, answer (b) must be correct since stratification of a random sample enhances this aspect of the research. The other methods are widely used, as discussed on pages 187 to 190. They are each types of ‘non-probability’ sampling which means the respondents in the sample have been selected for particular reasons and are therefore biased. This does not mean they are somehow invalid. On the contrary, they frequently offer insights into social behaviour that could not otherwise be obtained.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 187-190

‘Snowball’ sampling is employed most often when it is completely impossible to develop a sampling frame, as it was for Bryman’s own Disney project (see chapters 24 and 25 for the actual data and analysis). “Research in focus 8.4” gives an example of producing a sample of drug users by asking a few respondents to name others who might be interviewed, who in turn mention others, and so on. Although this sample-building technique is more likely to be used in qualitative research for purposes of induction, it can be used to quantify relationships among sample members, for example, within quantitative research.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 188,189

The primary reason for using a multi-stage cluster sample is geographic dispersion of the population. This automatically involves considerably extra time and money spent on travelling to conduct the interviews or surveys. However, if you select a sample on a more local basis you will not be able to extrapolate your results to the entire population. The solution is to select regions at random, for example, in the first stage, followed by cities, perhaps, as a second stage and local council areas as a third stage. In other words, by using this ‘multi-stage’ approach, we select ‘clusters’ of the national population at random, which can produce samples more easily studied.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 179, 180

Since ‘quota’ sampling is a type of ‘non-probability’ sampling, random selection cannot be one of its characteristics. It is somewhat less than scientific in its approach but can be very useful in providing quick indicators of response to events, which could later be tested on a probability sample. The researcher chooses respondents who are members of particular strata of society until a specified quota is reached. The quotas themselves are usually intended to reflect the size of the segment in the population as a whole.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 188-190

Once we know the size of the population to be researched, we can determine the size of our sample. This latter number will depend a lot on our resources of time and money. Then we make (or obtain, if one is already available) a sampling frame, from which we select our future respondents, typically using random number tables. This is to ensure that each member of the population has an equal chance of being selected, so there can be no bias in the selection, the result being referred to as a ‘simple’ random sample. If you answered (a) you were probably thinking of a ‘systematic’ sample, a short-cut method of selecting directly from the sampling frame but you must be careful to make sure the frame has not already been ordered in a particular way for another purpose.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 176,177

The findings of research based on random sampling of the population can be fairly applied to the population as a whole, but only to that population. This means that we must be very clear about the population we wish to study before drawing down the sample. There may be superficial resemblances between various populations but there may be substantial differences as well. We simply don’t know until we do the research. It is better to claim for your findings only that which can be defended, because this will earn greater respect for you and your work.
Reference: Bryman: Social Research Methods: 5th Edition Page(s) 193, 194