Return to MAIN (index) page Return to ASPECTS of LEARNING page Research - sampling




The group of people who are the subject of a piece of research is known as the "population".  (Note that this does not necessarily refer to the entire population of a country, or of a geographical locality - although it might.)



For example; in a piece of research into Year 6 pupils' perceptions of secondary school, the "population" would be ALL Year 6 pupils.





However, because it is often not feasible to investigate the entire "population" - for reasons of cost, time, accessibility, etc. - researchers are often obliged to obtain data from a smaller group, known as a "subset" of the population.  Researchers (usually) seek to define the subset in such a way that the data obtained will be representative of the total population.  This requires considerable care to ensure that the subset is not chosen in such a way that respondents' bias will affect the results of the research.



In our example, it would not be appropriate to choose a sample that was taken entirely from Year 6 pupils from inner-city primary schools.  Their perceptions of secondary school would most probably be based on the schools they know (which are likely to be inner-city schools) and would probably not be representative of the whole population (of Year 6 pupils).





It is not possible to say with any certainty what size sample is appropriate.  This will depend on the size of the population and the purpose of the research.  If the researcher plans to subject the gathered data to some kind of statistical analysis, the sample size needs to be large enough for that analysis to be meaningful.  Consequently, the researcher will need to consider the number of variables that might affect the heterogeneity (diversity) of the population.


When undertaking a survey, researchers are likely to prefer a large sample, to ensure that the effect of any variables within the sample are not exaggerated.  In qualitative research, however, the sample size is likely to be smaller - because of the need to gather more in-depth data.


Sample size is also likely to be constrained by cost, time, personnel and resources.





To ensure the validity of the research, the researcher must do all they can to ensure that the sample consulted represents the whole population in question.





Researchers need to bear in mind not only the practicability of the chosen sample - but also whether the sample can be accessed. 



Gaining access to Year 6 pupils, for instance, could pose all kinds of difficulties.  If it is planned to carry out interviews, issues are raised regarding parental permission, school co-operation, local education authority involvement, etc.  If it is intended to use questionnaires to gather data, not only does the researcher need to consider the practical matters of distribution and collection - but also the more fundamental issue of whether the questions will be properly understood, interpreted and answered by the target sample.





The two main methods of sampling are PROBABILITY SAMPLING and NON-PROBABILITY SAMPLING.




(also known as RANDOM sampling)


The chances of members of the wider population being selected for the sample are KNOWN.


Every member of the population has an equal chance of being included in the sample.



(also known as PURPOSIVE sampling)


The chances of members of the wider population being selected for the sample are UNKNOWN.


Some members of the population will definitely be included and others definitely excluded - and hence every member of the population does NOT have an equal chance of being included in the sample.  The researcher will deliberately select a particular section of the population for inclusion or exclusion (based on a justifiable rationale).





Most useful if a researcher wishes to make generalisations - because it seeks to be representative of the whole population.  There will be less risk of bias (whereas a non-probability sample, because it is not representative of the whole population, may betray bias.)


There are various types of probability samples:


Simple random sampling

A whole list of the population is drawn up and members are selected at random until the sample size is met.  (One problem associated with this method is that a complete list of the population is needed - and this may not be available, depending on the nature of the research and the population identified.)


Systematic sampling

Similar to simple random sampling.  A whole list of the population is drawn up and every nth member of the list is chosen in order to achieve the sample size.  (For example, if a sample of 100 is required from a population of 1000, every tenth member on the list is chosen - starting from a random point.)


Stratified sampling

The population is divided up into groups with similar characteristics (for example, males and females) then members are selected randomly from within these groups at the judgement of the researcher.


Cluster sampling

When the population is large and widely dispersed (such as Year 6 pupils), obtaining a random sample can pose administrative nightmares.  The researcher may opt to choose the sample from only a limited part of the population (say, from schools within a specific geographical area).


Stage sampling

This is an extension of cluster sampling.  The researcher may select the sample in stages - taking samples from samples.  (For example, having limited the choice to schools from a particular area, the researcher may then randomly choose only a limited number of schools from those available - and then only a limited number of pupils chosen at random from within the selected schools.) 


Multi-phase sampling

Whereas in stage sampling there is a single purpose throughout the sampling, in multi-phase sampling the purpose may change at each stage.  (For example, having limited the initial choice to schools from a particular area, the researcher may then choose a limited number of schools within the area based on whether they were inner-city or suburban - and then choose a limited number of pupils from within those schools in order to get a range of pupils of differing abilities; based, for instance, on their predicted KS2 SATs results.)





The usefulness of this approach derives from the fact that the researcher is able to target a specific group (in full knowledge that it does NOT represent the wider population).  This approach may be adequate if the researcher does not intend to generalise their findings.  It can also be particularly useful if the researcher is trying to determine why a specific section of the population appears not to be conforming to expectations.


Non-probability samples are easier to set up, less complicated to administer and can be especially useful for piloting a questionnaire prior to wider use amongst a wider population.


There are various types of non-probability samples:


Convenience sampling

The researcher simply chooses respondents from those closest to hand until the sample size has been obtained.  Also known as accidental or opportunity sampling.


Quota sampling

Similar to stratified sampling.  The population is divided up into groups with similar characteristics (for example, males and females) then members are selected randomly from within these groups - but the numbers selected are in proportion to their occurrence within the whole population


Purposive sampling

Researchers hand pick the members to be included in the sample on the basis of their typicality or specific characteristics.


Dimensional sampling

This is a simplification of quota sampling employed to reduce sample size.  The researcher identifies various factors of interest within the population (for example, it may be important to include the responses of people of different ages) then ensures that the sample includes respondents from each of the groups thus identified.


Snowball sampling

The researcher identifies a small number of respondents who possess a specific set of characteristics of interest.  The researcher then uses information provided by these respondents to get in touch with others who also possess the characteristics set.  This can be particularly useful for sampling a population where access is difficult - or who may not readily be identified by more conventional means (for example, gang members or drug addicts).




Cohen, L., Manion, L. & Morrison, K. (2000) Research Methods in Education. London. Routledge Farmer.




Click here to jump to Wikipedia article.