..Sample is used to represent a part of population as an entire population when you are
doing a research. It is can be selected for the research as a subset of population. And
the sample can be divided into a probability sample and a non-probability sample.
‘Probability sample is a sample that has been selected using random selection so that
each unit in the population has known chance of being selected’, Alan Bryman(2011)
claimed, ‘It is generally assumed that a representative sample is more likely to be the
outcome when this method of selection from the population is employed’. In his
opinion, maintaining sampling error to minimum is the probability sampling’s
purpose. For non-probability sample, it is the sample which has not been chosen by a
random selection approach. While probability sampling is the preferred approach,
however, in commercial studies, such as market research and case studies, this is
often very difficult to implement, in other words, this is even possible. In addition, it
often exceeds the researcher's budget, time and personal resources. So your sample
must choose the other way. Therefore, in many cases, the researchers will take
non-probability sampling. As for Mark Saunders, these methods are generally more
convenient for investigators, but the results of studies using such samples need to be
interpreted with caution. All non-probability sampling methods may limit the results
from the data. On the other hand, we can not deny that these methods can produce
reliable data when the researchers collect and interpret the inferences under the right
conditions. Therefore, you can specify the probability that any case will be included
in the sample. (Mark Saunders, 2003) Non-probability sampling provides a range of
alternative techniques based on your subjective judgment. In some exploratory
phases of research projects, such as pilot surveys, non-probability samples may be the
most practical, although it does not allow to determine the extent of the problem.
This essay would discuss the main benefits and limitations of using non-probability
samples in management research as this essay have already mentioned a little bit of
it in previous paragraph. Then this essay would have a critical discussion of when and
where non-probability samples are most appropriate and why in both academic and
professional research. Moreover, this essay would illustrate some potential issues
and implications of using non-probability for research design.
In non-probability sampling types, convenience sampling is the most common one.
Convenience sampling means that investigators will be more easily to find members
of groups they need, which is the opposite of probability sampling. This requires
investigators to find each total unit as a sample member at a given time and in the
environment. Cochran had a famous example, researchers will interview the people
in the crowded places such as shopping malls, relatively large commercial streets, and
so on. In these cases, the researcher will respect the wishes of the people to choose
whether they participate in the interviews. Although some of the survey respondents
were unhappy and troublesome, it was really convenient to use the resource sample as
a survey sample. In Mark Saunders’ opinion, for the convenience sampling, the
behavior of the individual determines whether they become the part of the study
sample, and therefore the deviation of results and the interpretation of the data(Mark
Saunders,2003). Moreover, Cochran (2016) claimed that because of the impact of
accident factors, the sample’s representation is too large to be guaranteed. Although
this sampling technique is widely used, it is prone to deviations and effects that are
uncontrollable because they are only present in the sample because they are readily
available. Typically, the sample is intended to represent the total population, for
example, the manager will be the MBA course as an agent for all managers! In this
case, the sample selection may be biased towards the sample, which means that the
following summary may be the most defective. These problems are less important in
situations where there is little variation in the population, and such samples are
commonly used as a test for studies using more structured samples. Lucas (1997)
describes an undergraduate study in order to understand the extent and variety of
student part-time work. Data were collected in the spring of 1995 from five students
at seven of the City University of Manchester City, and the authors were lecturers.
Self-administered questionnaires were distributed to students in different years.
Subjects were selected to maximize the number of breeds in the degree course type
and to provide a similar number of males and females. The questionnaires are issued,
completed and collected at the end of the classroom contact time by one of the
researchers or teaching staff. Kervin commented that ‘These programs represent a
very good attempt to produce different samples. This is a convincing sample because
the choice of a degree program is a selective choice, rather than a random choice,
because the absence of the classroom cannot answer the questionnaire. On the other
hand, who manages the questionnaires. An interesting question is whether
absenteeism may be connected to part-time work in some way; in other words, may
be absent because students work in class, or students may be too tired to attend
classes because of their part-time work?'(Kervin,1993)While these shortcomings are
unavoidable, sometimes convenience samples may be the only realistic sampling
strategy, especially in commercial research. A number of good studies have been
carried out using convenient samples. At the same time, we would understand that no
matter what specific research is, its results may be the illusion of the sample. After
all, the growth of science depends on a set of knowledge built on previous research.
This is particularly true for convenience samples, and the results of individual studies
should be examined on the basis of existing and future studies.
Quota sampling is a completely non-random sampling method that can be described
as classifying population according to certain criteria; then get the sample by
proportion of the number of sample per layer and the total number of layers (de Vaus,
D.A., 2002). In addition to the selected marker, quota sampling is usually used for
interview surveys. Quota sampling has many advantages over probability techniques.
In particular, it cost much less and can be set up very quickly. Barnett(1991)
illustrated that quota sampling is the only possibility if your data collection needs to
be done quickly, as with a television audience research survey. From 1st of April to 5th
of June, Ipsos MORI on behalf of Glasgow City Council conducted the 2016 Glasgow
family survey result. The topics were about quality of life and local resilience; local
environment; satisfactory services; board reputation and communication; financial
challenges affecting the Council; financial management; and equity and equity.
During this period, a sample of 1,023 Glasgow residents (aged 16 and above) was
interviewed for representative quotas. All interviews were conducted face-to-face
interviews in the respondents' homes using computer-assisted personal interview
(CAPI). In addition, it does not require a sampling frame, so it may be the only
technology that can be fully qualified if there is no other methods to choose from.
Once you have given another interviewer their specific task, they decide who will
interview until they complete the quota. You can combine the data for this job with
the data collected by other visitors to provide a complete sample. Because the
interviewer can choose your quota sample within the quota boundary of his or her
interview may be biased. Interviewers tend to choose respondents who are readily
available and willing to answer the survey.
Healey(1991) expounded that purposive sampling allows the researcher to select a
sampling method from a population that has been judged to be the most representative
population. When researchers are familiar with their field of study, it can represent
higher-grade samples. This sampling method can be applied to situations where the
population size is small and the internal variability is large, when the population
boundary can not be determined or the researchers’ time and labor and material
resources are limited. You can use your judgments to select the people who is best to
answer your research questions and meet your objectives. This type of sample is
usually used when very small samples are used, and the limited sampling can also be
used by the researcher’s theoretical approach. For such studies, the results of the data
collected from your initial sample inform you of the way the sample is extended to
subsequent cases (Section 12.6). However, such samples can not be considered
statistically representative of the total population.
If the researchers try to use the snowball sampling, they should has the initial
investigation needs of those who have the characteristics of the investigation needs,
and then the investigators will try to make the original of these people to provide
more qualified survey respondents, and then the third batch of respondents provide
the information for the next group of respondents, and so on. Mark Saunders
(2003) explained that the sample is just like a snowball rolled from small to large.
Snowball sampling is often used when you are difficult to identify the members of
the desired population or the general units of observational studies such as those who
work in claims for unemployment benefits. In snowball sampling, the primary
problem is initial contact. Once you finish this, these cases can identify more people,
and then identify other members, so the scope of the sample will be more widely.
However, some of these respondents cannot be found at the end of the sample, and
some molecules are missing in the supplier's mouth, both of which can lead to errors.
Since the representative problem for such a sample is immense because respondents
are most likely to identify other potential respondents who are similar to themselves.
The next question is to find these new cases. However, it is possible to provide a
unique possibility for a population that is difficult to identify snowball samples.
Self-selection sampling is the method that non-fixed, time and space continuous,
Cochran’2012’claimed, for instance, parades and assemblies have no exactly
population, participants move from one place to another, some participants leave
while some new participants come in, but these events occur within a certain range.
At the same time it is important to get the sample from the populations, as these
limitations which this essay have mentioned at previous paragraph earlier make it
impossible for the ensemble of samples to undergo too much temporal and spatial
variation. Specifically, many researchers separated by a certain distance, starting from
a certain direction, access to the nearest person, and then a number of steps to
investigate. Kervin(1992) provided that when you allow cases (usually individuals)
to identify their desire to participate in research, self-selected sampling occurs. You
therefore need to advertise the cases you need, by advertising through the
appropriate media or asking them to participate. The data is then collected from
respondents. Normally, the researchers choose the self-selection sampling because
they have feelings or opinions about research questions or stated goals. In some cases,
such as Adrian and his colleagues on the active management of redundant research,
which is what researchers they want. In this study, one letter from Personal Trade
News is significant for them because it produced a list of organizations which are
interested in their case.
In conclusion, non-probability sampling has the advantages of simplicity, low cost
and time saving as its main benefits. Therefore, it is used very widely in commercial
research and it has been completed a lot of good studies. However, this essay have
illustrated its main limitations. Since the subjectivity of the sampler cannot be ruled
out, and the representation cannot be controlled and objectively measured, the
sample does not have the properties of the inference population as a whole. These
limitations could be called potential issues of non-probability. Consequently,
non-probability sampling is mostly used for exploratory and preparatory studies, as
well as studies that are difficult to implement probabilistic sampling.