SAMPLING
INTRODUCTION
One of the most significant issues investigators have to consider when designing a project is the type and number of the people who will be included in the study. In this context, they have to consider a number of very important questions, such as:
· Will the whole population or a sample be studied?
· If sampling is preferred, which sampling procedure is most suitable?
· How large should the sample be?
· Is a sampling frame required?
· If so, is one available?
· How representative should the sample be?
· How will possible problems, errors and distortions be prevented?
· What kind of administrative arrangements are required for the selection of the sample?
· Are the required time, funds and staffing available, and if so, how can they be rationally employed?
· How will non-response be dealt with in the study?
· Are there any issues of ethics and objectivity to be considered at this stage, and how will such requirements be met?
The answers to these questions are many and the options diverse. One option is complete coverage of the population (saturation survey), whereby all units of the target population will be studied. In this case the target population is also the survey population. Another option, and the most common, is sampling, whereby the target population is investigated by studying a small part of it, namely a sample.
| Some basic concepts of sampling · Target population: The population for which information is required. · Survey population: The part of the target population that is studied. · Sample: The part of the survey population that is to be studied. · Sampling: The procedure employed to extract samples for study. · (Sampling) Units: The persons, groups, systems etc. chosen to be studied. · Saturation survey: A survey that includes all units of the target population. |
1 REASONS FOR SAMPLING
| Why use samples? · Necessity: In many cases a complete coverage of the population is not possible. · Effectiveness: Complete coverage may not offer substantial advantage over a sample survey. On the contrary, it is argued that sampling provides a better option since it addresses the survey population in a short period of time and produces comparable and equally valid results. · Economy of time: Studies based on samples take less time and produce quick answers. · Economy of labour: Sampling is less demanding in terms of labour requirements, since it covers only a small portion of the target population. · Overall economy: Sampling is also thought to be more economical, since it involves fewer people and requires less printed material, fewer general costs (travelling, accommodation etc.) and of course fewer experts. · More detailed information: Samples are thought to offer more detailed information and a high degree of accuracy because they deal with relatively small numbers of units. |
Problems:
· Sampling requires more intense and complex administration, planning and programming than saturation surveys.
· Sampling implies a reduction in the size of the target population and, hence, fewer potential respondents; this raises questions regarding representativeness and generalization of the findings that cannot be ignored.
2 PRINCIPLES OF SAMPLING
| Principles of sampling · Sample units must be chosen in a systematic and objective manner. · Sample units must be easily identifiable and clearly defined. · Sample units must be independent of each other, uniform and of the same size and should appear only once in the population. · Sample units are not interchangeable; the same units should be used throughout the study. · Once selected, units cannot be discarded. · The selection process should be based on sound criteria and should avoid errors, bias and distortions. · Researchers should adhere to the principles of research |
3 TYPES OF SAMPLING
Sampling procedures vary considerably. Samples may be constructed through self selection (respondents decide to take part in a study, for example, in response to media calls for volunteers) or, as is most common, through the researcher. There are also sampling procedures based on probability standards (random or probability samples), and on non-probability standards (non-probability samples).
Criteria of probability and non-probability samples
| Probability sampling | Non-probability sampling |
| · Employs probability theory · Is relatively large · Size is statistically determined · Size is fixed · Sample is chosen before the research · Controls researcher bias · Involves complex procedures · Has fixed parameters · Involves high costs · Planning is time consuming · Is designed to be representative · Planning is laborious · Treats respondents as units · Facilitates inductive generalizations · Is employed in quantitative research | · Does not employ probability theory · Is small, often covering a few typical cases · Size is not determined statistically · Size is flexible, but can also be fixed · Sample is chosen before and during the research · Does not control researcher bias · Involves simple procedures · Has flexible parameters · Involves relatively low costs · Planning is not time consuming · Representativeness is limited · Planning is relatively easy · Treats respondents as people · Facilitates analytical generalizations · Is mostly for qualitative research |
3.1 Probability (random) sampling
Probability sampling is the procedure in which the choice of respondents is guided by the probability principle, according to which every unit of the target population has an equal, calculable and non-zero probability of being included in the sample. There are several forms of probability sampling, but simple random sampling and systematic random sampling are the most common.
Simple random sampling
The characteristic of this type of random sampling is that the sampling units, apart from having an equal chance of being selected, are independent from each other. Their chance of being selected does not depend on the selection of other units. The three most common methods of simple random sampling are: the lottery method, the random numbers method and the computer method.
The lottery method: Choosing respondents by the lottery method entails a procedure that can be described as follows:
Step 1 Identify or construct a sampling frame, that is, a list of the units of the target population. Such frames may for instance be the electoral role, student records, rating records or similar lists. Include the names and addresses of sample units in alphabetical order and numbered accordingly.
Step 2 Determine the sample size, that is, the number of units required for the study.
Step 3 Place a number of small discs or balls in a container, numbered to correspond to the names contained in the sampling frame. If 500 names are listed in the frame, there should be 500 balls or discs in the urn, numbered from 1 to 500.
Step 4 Mix well and remove one ball from the urn. The number of this ball is registered and the corresponding name in the sampling frame is noted." This is the name of the first respondent. The ball is either returned to the urn or left out; either method may be employed. Continue this process until sufficient names have been selected. (If an already drawn number is selected it is ignored). The selected respondents constitute the sample.
The random numbers method: This method is similar to the lottery method, except that the urn and balls are replaced by random number tables, which are available in separate publications or in the appendices of statistics texts. Choosing the sample by using the random numbers method involves the following steps:
Step 1 Identifying or constructing a sampling frame.
Step 2 Selecting appropriate tables of random numbers.
Step 3 Picking numbers from the tables randomly and registering them; the names in the sample frame that correspond to these numbers constitute the sample.
The computer method: In this method, we instruct the computer to give us a set of numbers equal to the number of sample units, for example, 500 numbers ranging between 1 and 6,000. Having the numbers, we follow the model employed in the previous methods. This technique is employed when the sampling frame is not in the computer; but if it is electronically available, we instruct the computer to choose, say, 500 names from the list with a simple command. Obviously the computer requires a small program to complete this task.
Systematic random sampling
Systematic random sampling is a procedure in which the sampling units are not only chosen randomly – as in simple random sampling – but in which this random choice is also integrated with the choice of another sampling unit. This is what distinguishes this type of sampling from simple random sampling, and what qualities it to be termed systematic. The actual choice of units is arranged through calculation that aims, first, to maintain randomness in selection, and second to spread the sampling units evenly throughout the list of respondents (the sampling frame). The system is based on the sampling fraction method.
In this method, units are drawn from a sampling frame by means of the sampling fraction (symbolized by k) that is equal to N/n, where N is the number of units in the target population and n the number of units of the sample. For instance, if the target population is 4,800 and the intended sample size 600, the sampling fraction is 8 (k = 4,800/600 = 8). To select a sample by using the sampling fraction method, we proceed as follows:
Step 1 Identify or construct a sampling frame.
Step 2 Determine the sample size.
Step 3 Calculate the sampling fraction k (as above, k = N/n).
Step 4 Randomly select a number between 1 and k. In the above example, since k = 8, the random number would be between 1 and 8; let us say 6.
Step 5 Record the random number (6) and every eighth number after 6, unit 6,000 is reached, e.g. 6, 14 (6 + 8 = 14), 22 (14 + 8 = 22), 30 (22+8 = 30) etc.
Step 6 Locate the names in the sampling frame that correspond to the selected numbers.
The respondents thus identified constitute the sample.
Stratified random sampling
Stratified random sampling is a probability sampling procedure in which the target population is divided into a number of strata, and a sample is drawn from each stratum. The resulting sub-samples make up the final sample of the study. The strength of this procedure is in that it allows all population groups to be represented in the final sample. The division of the population into strata is based on one or more significant criteria, such as sex, age, ethnic background, race or economic status.
The sample size can be proportionate or disproportionate to the units of the target population. This means that the samples taken from each stratum can be either proportional or disproportional to the size of the samples. As indicated above, a stratified sample is employed when there is a need to represent all groups of the target population in the sample, and when the researcher has a special interest in certain strata'. In this sense, the method, is very economical, and offers a high degree of representativeness. A stratified sample is drawn as follows:
Step 1 The target population is divided into a number of strata, according to the number of the significant groups in the population.
Step 2 The sampling frames for each of these groups are identified; if these are not available, relevant frames must be developed.
Step 3 Employing one of the methods discussed above, a sample is drawn from each group. This can be proportionate or disproportionate to the number of units in the population.
Step 4 The individual samples are merged into one; this constitutes the sample for the study.
Multi-stage sampling
In multi-stage sampling, the selection of sample units begins with the choice of a large sample, and proceeds with new samples taken in succession from those previously selected, thus facilitating the construction of a more suitable and more effective choice. More specifically, a large sample is chosen, using a random sampling procedure, and then another sample is taken from within this sample, excluding excess and unrelated units. For instance, if the study is to focus on professional women, all men and non- professional women contained in the first sample will be discarded. If required, another sample is chosen from the second sample for similar reasons. This process is continued for as long as required, with each additional drawing making the sample more specific, more focused, more relevant to the research object, and more representative. The characteristic of this type of sampling is that data collection is conducted only from the final sample.
The process of choosing a sample through the multi-stage sampling method proceeds as follows:
Step 1 A sampling frame for the target population is identified.
Step 2 A large probability sample is chosen; the units of this sample are usually referred to as primary selection units. A sample from the primary selection units is then chosen.
Step 3 After the criteria of the respondents have been identified (in terms of gender, ethnicity, marital status etc.), another sample is drawn from within this sample. In most cases, a second sample is sufficient to meet the requirements of the study. Otherwise, the procedure is repeated until the targeted sample size is reached.
Step 4 The final group constitutes the sample of the study.
The use of several screenings and drawings is not only time-consuming but also expensive. Hence, multistage sampling procedures are employed only when absolutely necessary. An obvious case for such a sampling procedure is a heterogeneous population where there is not enough information to permit the construction of a representative sample without screening.
Cluster (Area) Random Sampling
The problem with random sampling methods when we have to sample a population that's disbursed across a wide geographic region is that you will have to cover a lot of ground geographically in order to get to each of the units you sampled. Imagine taking a simple random sample of all the residents of New York State in order to conduct personal interviews. By the luck of the draw you will wind up with respondents who come from all over the state. Your interviewers are going to have a lot of traveling to do. It is for precisely this problem that cluster or area random sampling was invented.
In cluster sampling, we follow these steps:
§ divide population into clusters (usually along geographic boundaries)
§ randomly sample clusters
§ measure all units within sampled clusters
Example: Let's say that we have to do a survey of town governments of the counties in New York. That will require us going to the towns personally. If we do a simple random sample state-wide, we'll have to cover the entire state geographically. Instead, we decide to do a cluster sampling of five counties. Once these are selected, we go to every town government in the five areas. Clearly this strategy will help us to economize on our mileage. Cluster or area sampling, then, is useful in situations like this, and is done primarily for efficiency of administration. Note also, that we probably don't have to worry about using this approach if we are conducting a mail or telephone survey because it doesn't matter as much (or cost more or raise inefficiency) where we call or send letters to.
Multi-phase sampling
The sample selection within this procedure is identical to multi-stage sampling, with the difference that in this sampling procedure, each sample is adequately studied before the next sample is drawn. This offers an advantage over other methods, because the information gathered at each phase helps the researcher to focus the selection more effectively and more constructively in later phases.
Spatial sampling
This procedure is employed when the study addresses people temporarily congregated in a space, and the data have to be collected before the crowd is dispersed. An example of such cases is the study of the views of people demonstrating in a city square about tax policies. Due to the nature of the population, there are neither sampling frames nor sufficient time available to permit the use of other methods. Apart from this, data collection has to be conducted so that a relatively representative coverage is achieved, randomly and in a systematic way before the crowd disperses. The way this is usually done is shown in the following example.
The details of the procedure can be changed to meet the actual circumstances of the situation. For instance, if a crowd is large, the interviewer might have to walk a longer distance, say ten steps. Also, it may be necessary for the interviewer to seek specific people; for example, the first will be male, the second female, and so on, or to include a variety of respondents (young, old, ethnic, non-ethnics and so on).
3.2 Non-probability sampling
As the name indicates, non-probability sampling procedures do not employ the rules of probability theory, do not ensure representativeness, and are mostly used in exploratory research and qualitative analysis. Some of these techniques can, with some adjustment, be converted into probability methods. Accidental sampling, purposive sampling, quota sampling and snowball sampling are examples of non-probability sampling techniques; they are presented below.
Accidental sampling
This procedure employs no systematic techniques to choose the respondents. Instead the simple units are those people who 'accidentally' come into contact with the researcher, for instance, the researcher may stand at a street-corner, in front of a school or church, or at the main exit of a shopping centre, and ask a number of people passing by to take part in the study. They are chosen 'by accident' – they just happen to be there at that time – hence the name of the sampling procedure (there are several other names for it, including 'convenience sampling 'chunk sampling', 'grab sampling' and 'haphazard sampling').
The researcher here is not interested in representativeness, objectivity, validity or similar considerations, but in getting information that would reveal certain aspects of the lifestyle in question and in certain cases give information about typical cases.
Purposive sampling
In this technique the researchers purposely choose subjects who, in their opinion, are relevant to the project. The choice of respondents is guided by the judgment of the investigator. For this reason it is also known as judgmental sampling. There are no particular procedures involved in the actual choice of subjects.
In such cases the important criterion of choice is the knowledge and expertise of the respondents, and hence their suitability for the study.
Quota sampling
Quota sampling is a procedure in which the researcher sets a 'quota' of respondents in to be chosen from specific population groups, defining the basis of choice (gender, marital status, ethnicity, education etc.) and determining its size (e.g. 60 parents of toddlers; 35 policewomen; 66 teachers and so on). The choice of the actual respondents is usually left up to the interviewer.
More specifically, the researcher considers all significant dimensions of the population and ensures that each dimension will be represented in the sample. This is usually referred to as dimensional sampling, and is particularly useful when the sample is small. In such cases, this procedure guarantees that at least one case from each dimension of the population will be included in the sample.
Quota sampling is quite common in the social sciences because it is less costly than other techniques, does not require sampling frames, is relatively effective, and can be completed in a very short period of time. It is limited, however, especially with respect to representativeness, control of sampling and fieldwork requirements, which in such studies are not relevant. It must be noted, however, that the choice of the respondents can be determined more strictly, by employing probability rules, hence requiring sampling frames and specific methods of selection, while retaining the quota factor. This would convert' the procedure to a probability sampling, which is quite possible.
Snowball sampling
In this approach, the researcher chooses a few respondents using accidental sampling or any other method, and asks them to recommend other people who meet the criteria of the research and who might be willing to participate in the project. This process is continued with the new respondents until saturation - that is, until no more substantial information can be acquired through additional respondents – or until no more respondents are available.
This method is employed when the lack of sampling frames makes it impossible for the researcher to achieve a probability sample, when the target population is unknown, or when it is difficult to approach the respondents in any other way. In many cases, snowball sampling is the only way of securing a sample for a study.
Theoretical sampling
Theoretical sampling is not a form of 'sampling' in the sense we introduced it above. In the words of its creators, theoretical sampling is the process of data collection for generating theory'. The focus here is on data collection rather than on the choice of respondents.
In very simple terms, in theoretical sampling the sample units are not simply 'chosen' by the researcher prior to the commencement of the study but determined by the knowledge that emerges during the study. The researcher chooses the first respondent, collects relevant information and knowledge about the research topic, and on the basis of this decides which person to study next. The direction of 'theory' that develops during the research process determines who the next respondent will be, a decision that could have not been made at the start of the study. Theoretical sampling is interconnected with data gathering and serves to enable comparisons in time and place so as to discover variations in concepts and to integrate categories in terms of their properties and dimensions.
In theoretical sampling, the study does not continue until all respondents have been contacted but rather until the process of study indicates that saturation has been reached; that is achieved when data collection no longer generates new data, when the categories are well developed in their properties and dimension', and when the relationships among these categories are 'established and validated'.
| The nature of theoretical sampling Theoretical sampling: · Entails an ongoing process: The sample here is not chosen and fixed at the outset and before the commencement of research, but chosen in an ongoing process that goes right through the whole study. · Involves 'places, people and events': Theoretical sampling takes into account the fact that people think and act differently, depending on many factors. Taking these factors into consideration helps the researcher to test, verify and contrast emerging concepts, categories and theory. · Is guided by the emerging theory: Theoretical sampling is self-regulated in the sense that it guides data collection and analysis towards developing a theory, which in turn directs the nature and content of sampling. · Is concerned with developing and validating theory: It is geared towards assuring that the emerging theory is adequately tested, so that it can be granted validity and high quality. · Ends when theoretical saturation has been reached: It is not the identification of the respondents but the completion of the research that brings sampling to an end. (Benini, 2000) |
4 SAMPLING PROCEDURES IN QUALITATIVE RESEARCH
Qualitative researchers employ sampling procedures that correspond to the philosophy of this type of research, and that are less structured, less quantitative and less strict than the technique quantitative researchers employ.
Normally, qualitative studies employ a form of non-probability sampling, such as accidental or purposive sampling, as well as snowball sampling and theoretical sampling. Qualitative sampling is biased by the nature of the underlying qualitative framework, which is perceived as an investigative process
Nevertheless, qualitative research has no strict, agreed rules for sampling employed by all researchers. Sampling procedures employed by qualitative researchers include those mentioned above (accidental, purposive, snowball sampling and so on) or a version or combination of quantitative sampling procedures. In all cases, sampling is closely associated with theory. It is therefore either theory-driven 'up-front', where subjects are chosen before data collection, guided by theory, or progressively, during data collection. The latter is known as theoretical sampling, and is connected with grounded theory.
Irrespective of the type of sampling chosen, several sampling parameters must be considered before a qualitative study can begin. Although qualitative sampling is a function of the research process itself that is decided on while the research is in progress and depends on the outcome of the study, researchers do have to decide at the outset at least about a number of issues, such as the informants or respondents who will be studied, the selling where research will take place, the events and processes to be considered in the investigation, and the time when research will be conducted. In any case, sampling procedures in qualitative research are inevitably related to a number of issues and choices, a few of which are listed below.
· Kind of people: The kind of people (actors) who will be included in their study, at least those to begin the study with.
· Time: The study may be conducted on working days, on weekends, during the school holidays, in summer or winter, in the afternoons, in the evenings or at any other time.
· Kind of event: The kind of event or processes to be studied; whether it will be a routine event, a special event, an unexpected event, or all types of events.
· Setting: The context in which the research will be conducted (e.g. the home, the club, the work place, a friend's house).
In summary, the sampling procedures employed by qualitative researchers demonstrate a number of characteristics; those presented in the following Box are considered by a number of writers to be most significant.
| Criteria of qualitative sampling Qualitative sampling is directed: · not towards large numbers of respondents but rather towards typical cases · not towards fixed samples but towards ones that are flexible in size, type or subjects · not towards statistical or random sampling but towards purposive sampling · not towards "mechanical" sampling but towards theoretical sampling · towards fewer global settings than quantitative sampling · not towards choosing a sample before the study has started, but (often) while the study is in progress · not towards a strictly defined size but a sample whose number will be adjusted while the study is in operation · not towards representativeness but rather towards suitability |
5 INTERNET SAMPLING
The popularity of and easy access to the Internet has affected the conduct of social research in many ways. This is clearly shown in sampling, which has begun to adjust its techniques to new ways to approach people in the community. More and more research bodies using it as their preferred sampling procedure.
In simple terms, Internet sampling is a procedure that is administered, partly or fully, through the Internet. The researchers bring questionnaires to the attention of prospective respondents, by either directly forwarding them the questionnaire, or informing them of the availability of the survey and asking them to participate. This is facilitated through email or web pages.
Email: Researchers who gain access to email lists act in one of two ways. They may send a message to the address, asking the email holder to volunteer and take part in the survey. This is usually conducted as a part of the usual spamming, uninvited email. The other approach is for researchers to attach the questionnaire to an email sent to all members of in email list, with an invitation to participate. In a number of cases, participation is associated with a variety of rewards.
Web pages (URL): The same procedure is employed when respondents are sought in the web page, where readers are asked to complete a questionnaire, and if they agree they are directed to the questionnaire. Internet users will come across the researcher's message or the advertisement of an advertising agency, or both, and will respond according to their interests.
In a sense, Internet recruiting of respondents is not very different from advertising a study in the media, or contacting people on the phone. As in the latter approach, Internet sampling has to deal with problems of representativeness; the number of Internet users is limited, and is significantly lower among older people. Researchers make an effort to overcome this weakness by enlarging the population basis of prospective respondents, for example by means of sample triangulation, but the problem remains. The extensive use of spamming and the resulting anger of Internet users at this annoying interference often make the response to uninvited invitations far from pleasant. As noted above, offering rewards for full participation may be one answer to the problem.
6 SAMPLE SIZE
The question about appropriate sample size in social research is given due attention by researchers of all schools of thought. However, the focus of relevant estimations varies significantly, with some showing an interest in pure quantity, others in quality and others again in both. A wise rule in this case is: the sample must be ‘as large as necessary, and as small as possible'. This critical figure is reached in some cases through logical estimates, and in others through statistical computation.
6.1 Non-statistical estimations
Sample size is directly associated with two major factors: the paradigm that guides the research, and the nature of the target population. These are the major determinants of the size of the sample, at least in logical terms. In quantitative research, both are seriously considered when the sample size is addressed. In qualitative research, the paradigm guides the process, but the nature of the data obtained will determine the size, and this is unpredictable. The study will stop when saturation is achieved, and this emerges out of the data and not out of logical thinking or other calculations. There are, however, qualitative researchers who follow the quantitative paradigm, and estimate their sample size in advance. Hence, the matter is not that simple!
Be that as it may, quantitative researchers and some qualitative researchers will come to a decision regarding the size of their sample before the study begins. The guiding factors in this context are associated with the type of population, the type of methodology employed, the availability of time and resources, the aim of the research, the type of instruments used, the accuracy required and the capacity of the research team. More particularly, the parameters are listed in the Box.
| Some determinants of sample size The size of the sample depends on the following: · Underlying methodology: Quantitative research requires larger samples than qualitative. · Nature of the study object: Some research topics require large and others small samples. · Available time and resources: · Homogeneity of the target population: The more homogeneous the target population the smaller the sample can be, and vice versa. · Accuracy: The higher the degree of accuracy required, the larger the sample. · Nature of the data required: If quantitative data are required, a large sample is needed; if qualitative data are required, a small sample will be sufficient. · Purpose of the study: If the study aims to achieve inductive generalizations, a large sample will be required. · Intensity of the study: The more intense and in-depth the method of data collection, the smaller the sample size. · Nature of the study: Surveys require a large sample; case studies do not. · Response rate: The lower the expected response rate, the larger the sample size. |
It is worth noting that large samples do not always guarantee a higher degree of precision, validity and success in general. The quality at the results depends on several factors and the sample size is only one of them.
6.2 Statistical computations
Many quantitative researchers employ statistical methods in order to define the 'right 'size of the sample. This is based on the assumption that if certain data are available, the sample size can be statistically computed so that sampling errors can be reduced to a minimum or to an acceptable or expected level. There are several methods employed by statisticians and social researchers, some of which are quite complicated and beyond the limits of this treatise.
In general, the logic of many statistical methods relates sampling error to the standard error (SE): if the standard error is reduced, the sampling error is also reduced. The standard error depends on the size of the sample: with increasing sample size the standard error is decreased. Thus an acceptable standard error can be achieved by changing the sample size. This method manipulates the size of a sample by increasing or reducing it, until it corresponds to a standard error that is considered acceptable. This is then the ideal sample size.
The standard error varies inversely with the square root of the sample. If for instance, we intend to reduce the standard error, we have to increase the sample size. Thus, if we wish to determine the sample size that will reduce sampling errors, we start with a sample size taken at random, compute the standard error and increase the sample size until the relevant standard error is at an acceptable level. This manipulation works well with small samples, where increases in sample size result in increases in accuracy (i.e. decreases in sampling error), but it does not work equally well with large samples. Above a certain point, the increase in size required to achieve a significant decrease in the error is so large and therefore so costly that it makes such an increase not worth the effort.
MAIN POINTS
· Sampling is the process of choosing the respondents and the units of the study in general.
· Sampling is a common practice and an indispensable research tool in social sciences.
· As the alternative to conducting a saturation survey, sampling offers many advantages.
· In quantitative research, sampling units are chosen prior to the commencement of the study, objectively and systematically; are easily identifiable and clearly defined; independent from each other; not interchangeable; and free of errors, bias and distortions.
· The two distinct types of sampling are probability and non-probability sampling.
· In probability sampling, all units have an equal, calculable and non-zero probability of being included in the sample.
· Non-probability sampling does not adhere to the rules of probability.
· Qualitative researchers employ non-probability sampling procedures such as theoretical sampling, accidental sampling, purposive sampling, quota sampling and snowball sampling.
· Usually, one sample is sufficient to conduct a study, but multi-sample studies are also common.
· Sample size is defined using statistical and non-statistical methods.
No comments:
Post a Comment