SURVEYS: POPULATION AND SAMPLE

Quantitative research is a widely used survey research method that aims to systematically gather information from respondents to predict and understand various aspects of the population's behaviour. This involves sampling, designing questionnaires, administering surveys, and analysing data. Different types of surveys include in-person and telephone interviews, omnibus surveys, and self-administered questionnaires.

Despite its effectiveness, quantitative research has some limitations. It primarily collects naturally quantitative data and analyses it using mathematical methods. However, researchers can employ measuring instruments to convert non-quantitative phenomena, such as beliefs, into quantitative data. Many researchers adopt a pragmatic approach and use quantitative methods to gain a broad understanding, test hypotheses, or study quantitative phenomena.

Survey Research has become a major, if not the primary, technique for studying and asking participants one or more questions about their attitudes, perceptions, or behaviours (Stockemer, 2019).

Survey research is a widely utilised quantitative method in social sciences, education, and healthcare. It entails gathering standardised data through a series of questions posed to participants, which can then be analysed statistically. Through this statistical analysis, researchers can conclude the broader population from which the sample was taken. Surveys are commonly used better to understand individuals' beliefs, attitudes, and opinions and to report their experiences and behaviours. While the data collected from surveys is subjective, more objective data such as age and place of birth can be obtained through other means (Leavy, 2022).

There are two primary methodological designs in survey research.

Cross-Sectional Survey is a method of collecting data about a group of individuals at a specific time (Mtshweni, 2019). Such surveys are often legitimate from a theoretical standpoint, as they allow researchers to conclude the relationship between independent and dependent variables. However, since only one data set is available for independent and dependent variables, cross-sectional surveys cannot prove causality. Researchers must rely on theory, logic, and intuition to support the findings and conclusions from cross-sectional studies. In other words, researchers should only use cross-sectional data to test theories if there is a clear temporal relationship between the independent and dependent variables. A cross-sectional survey can be a powerful tool for testing hypotheses when there are explicit theoretical assumptions about the relationship. Nonetheless, empirical relationships are only sometimes clear-cut, and it can be challenging to derive causal explanations from cross-sectional studies.

Longitudinal Surveys are a unique type of research that differs from cross-sectional studies in that they are conducted repeatedly over an extended period of time. Each survey consists of the same set of questions, allowing researchers to gain invaluable insights into the evolution of attitudes and behaviours within a population. These surveys can be classified into three categories: trend, cohort, and panel (Leavy, 2022; Stockemer, 2019; Watson, 2015).

• A Trend study, often called a repeated cross-sectional survey, involves conducting multiple surveys with different groups of individuals over time. The surveys include the same questions in each wave, fully or partially. This method enables researchers to identify significant shifts in attitudes and behaviours over time.
• Cohort studies have a narrower focus, as they focus on a particular group rather than the entire population. Like trend studies, cohort studies involve recurring inquiries directed solely at a selected group of people with a shared characteristic. Each iteration involves a new sample from the same population, indicating that the population remains constant while the individuals sampled vary (Price & Lovell, 2018).
• Finally, Panel studies are known for asking the same set of questions to the same individuals in multiple waves. While these surveys can be quite costly and challenging to conduct, they are considered the most effective way to detect causal relationships or changes in individual behaviour. As a result, panel studies serve as a powerful tool in identifying such relationships.

Sampling is a crucial aspect of research that enables researchers to collect data about a population of interest. The primary objective of sampling is to acquire data representative of the entire population, which can be challenging due to the population's size and diversity. To achieve a representative sample, researchers must ensure that the sample consists of individuals with the same characteristics as the population of interest. This can be achieved through various sampling techniques, such as random, purposive, and quota.

A representative sample consists of individuals with the same characteristics as the population. For example, suppose a researcher knows that 55% of the population he intends to study is male, 18% are African-American, 7% are homeless, and 23% earn more than 100,000 euros. In that case, he/she should try to match these characteristics in the sample to represent the population.

Random sampling is used when researchers cannot match the population's characteristics in the sample. Randomization helps to offset confounding effects by randomly selecting cases. Fig. 8. displays a graphical representation of a population (people, events, households, institutions, or something else) that is the subject of research, a sample frame (set of units from which a sample will be drawn: in the case of a simple random sample, all units in the sampling frame have an equal chance of being drawn and of occurring in the sample), and a sample (the subset of the population chosen for the research or survey).

A biased sample is neither representative nor random. Its answers do not reflect those obtained from the entire population. Survey responses can suffer from different biases, such as selection bias, non-response bias, and response bias. Sampling error is always present due to statistical imprecision.

Convenience sampling is a non-probabilistic sampling technique where people are chosen because they are readily available. In purposive sampling, subjects are selected based on predetermined characteristics. Volunteer and snowball sampling are other non-probabilistic sampling techniques employed in populations that are difficult to access. Quota sampling is a technique used in online surveys where sampling is done based on pre-established criteria. For instance, many polls have an implicit quota, such as customer satisfaction.

Selecting the optimal sample size is crucial for quantitative researchers seeking precise and accurate results from significance tests. Determining the sample size involves various methods, such as statistical formulas and electronic sample size calculators. However, researchers must first establish details about the target population to identify the appropriate sample size. They must consider important factors such as the population size, error level, confidence interval, and confidence level. These determinants aid in deciding how much the sample mean can deviate from the population mean and how sure researchers want to be that the actual mean will fall within their confidence range. The confidence interval is usually set at 90%, 95%, or 99% confidence.

Furthermore, researchers must also account for the standard deviation to anticipate the variation between responses. The size of the quantitative sample is assessed based on the strength of a hypothesis test and the quality of produced estimates (Mwansa et al., 2022). Five important study design parameters, such as the significance criterion, minimum expected difference, estimated measurement variability, desired statistical power, and one or two-tailed statistical analysis, typically determine the primary determinant of sample size in a quantitative design.

Probability sampling is a method employed in quantitative research designs that aims to ensure that each member of the population has an equal chance of being included in the sample (Mwansa et al., 2022). This method's primary objective is to ensure that researchers draw valid conclusions from their findings and that their results represent the entire population. Researchers employ four primary sampling techniques to achieve this, which we will discuss below.

Simple random sampling involves using completely random techniques or tools, such as random number generators, to give each individual in the population an equal chance of being selected.

Systematic sampling is like simple sampling, but specific individuals are chosen regularly. However, it is essential to ensure that the list does not contain hidden patterns that could skew the sample (Mwansa et al., 2022).

When sampling the population, stratified sampling involves dividing it into different subpopulations that are significantly different from one another. Each subgroup is well-represented in the sample, and researchers must divide the population into subgroups based on related characteristics, such as gender, age group, income class, or function. Then, they randomly or systematically select samples from each subgroup.

Cluster sampling involves dividing a population into subgroups with characteristics comparable to those of the sample as a whole and randomly selecting entire subgroups. This method is proper when dealing with large and dispersed populations, but it is more likely to introduce sampling errors as there can be significant differences between clusters.

On the other hand, non-probability sampling is a subjective approach to selecting units from a population, making it a fast, easy, and inexpensive way of obtaining data. However, it assumes that the sample is representative of the population, which can be a risky assumption. Additionally, elements are chosen arbitrarily, making it impossible to estimate the probability of any element being included in the sample or identify possible bias.

Convenience sampling, also known as random sampling, utilises individuals who are most easily accessible as study participants.

Snowball sampling, also known as chain sampling or network sampling, asks early sample members to find and refer additional people who meet eligibility requirements.

Quota sampling involves the researcher determining the necessary number of participants from each population stratum and identifying population strata.

Lastly, Purposive sampling, also known as judgmental sampling, is based on the idea that the researcher's understanding of the population can select individuals for the sample.