The sample is the people you will experiment on in your research to obtain results. In an ideal world the study should examine an unbiased representative mini-chunk of the actual population, so that the results can easily be extrapolated and applied to the real world. To get your sample as close to the real world as possible, you must consider:
- sample size
- who should be in your sample?
- the ethics of all this experimenting on the poor sample people
Who should be in your sample?
The more accurate the sample represents all the different people in the real population, the more the results can be applied to the real world and not be a waste of time. The more precise the method, the more time-consuming and expensive it usually is, so this needs to be taken into account depending how much time and funding you have spare. Member checking improves the validity of a small sample.
What is random sampling?
Random samples are when people are chosen at random from a population to be analysed. For example sticking a pin in the phone book-type thing. A cross-sectional study is when a sample of the population is observed or assessed (through surveys/questionnaires) at one point in time. This provides a snapshot of opinions and are useful for measuring attitudes and behaviours. However there are the following flaws always with cross-sectional studies (Aveyard 2014): a compete response to the questionnaires is v.unlikely, as is the sample is unlikely to be 100% representative, therefore the results will only ever be from a selection of but a random sample. People may be selected against because they don’t like questionnaires or don’t think the topic is worth responding to, so an incomplete picture is achieved.
A biased sample would be selecting OTs from the COT register, as not all OTs chose to be a member. An unbiased selction would be from HCPC’s database because this is a record of ALL qualified practising OTs.
What are the different types of purposive sampling?
Purposive sampling is when participants are chosen because of exposure to a phenomenon, the opposite of random sampling (Flick 2014). There are a number of different types, each with different characteristics (…and strengths & limitations) meaning there are reasons why you would chose one over another. General advantage is that purposive sampling will yield richer data by focussing analysis on those with the largest potential for rich data; but disadvantage is that by selecting for one type of experience the researcher may inadvertently exclude some relevant themes which would have been present if the subject range was broader (Rebar et al 2015). The key is not to have too narrow a sample to avoid this.
- complete collection
- certain criteria are put forwards in advance and everyone who meets that criteria is sampled. This would mainly be feasible in regional studies eg all the married 30-40yo men with prostate cancer in Berkshire hospitals. This isn’t helpful for developing new theories since at the outset the sample is already decided. Instead perhaps testing assumptions or comparing specific groups.
- sample is calculated beforehand to be an accurate representation of the real population, and the criteria are known data are collected.
- aka Glaser & Strauss (1967) grounded theory, where depending on the results of initial data analysis, guides where next data come from, to build up a theory. The initial sample isnt selected according to criteria or randomly, rather a group is selected because of their predicted ability to provide insights on your research area. Issue here is how to decide when to stop adding new data-or when theoretical saturation has been reached. This is when no new information is being added to the data…ie the same things you’ve already heard are being repeated in interviews.
- extreme case
- especially successful or relevant experiences are selected to be analysed, eg OTs who have been working for the longest in stroke rehab settings, or the newest qualified OTs. Or clients who have been the most successful in completing rehab, or those who have relapsed the most, etc.
- typical case
- choosing people who represent typical situations, eg those who recover from hip fractures and leave hospital in a usual 4-6 week timeframe. Provides information about the usual experience
- maximal variation
- chose people from either end of the spectrum from extreme case sampling- eg those newest qualified and those in practice for 20+years. Provides benchmarks for the limits of the experience or its occurance
- chosen for the intensity or prevalence that an experience occurs- eg someone who claims to never be late for work or someone reported as showing up late every damn day
- critical or sensitive
- people where the phenomenon to be studied will be extra clear- interviewing experts in the field for example
- people selected according to researcher knowledge, eg that there will be a load of students they know all had bad experiences on MH placement at a conference, so the researcher uses the students at the conference location as the sample.
- selecting people due to the ease of access in the circumstances, such as restricted resources or time. The RCOT and HCPC will neither release contact data for OTs and so it’s not possible to send out surveys to a random selection of all registered UK OTs, or those members of the professional body. Instead you may be forced to employ convenience sampling whereby you analyse the people you are able to reach through social media or contact with employer NHS trusts. Participants are included in the sample on a first come-first served basis (Rebar et al 2015). Advantages are this is inexpensive, so perhaps useful for an exploratory study which is determining IF there is any kind of relationship present. Disadvantage is that it wont provide as rich data as purposive sampling, as you may miss sources of data that arent as easily accessible, eg private practice OTs when you only recruit though NHS trusts.
- contacts from one relevant participant’s network are used to source further subjects (Rebar et al 2015). Convenient and also useful when it is difficult to identify subjects conventionally. However it risks focusing the research prematurely.