Resources
The Breakdown
Important
- Sampling and Sample Size in Research – It is fundamentally important for any study, as it defines the participants upon whom the research will be conducted.
- Sampling Frame – A critical component of the sampling procedure, where the specific characteristics, including inclusion and exclusion criteria, for study participants are precisely defined.
- Previous Studies for Sample Size Estimation – It is essential to consult prior research on similar topics to determine and justify the appropriate sample size for a current study.
- Quota Sampling vs. Stratified Sampling – You need to know that quota sampling is the non-random sampling method most similar to stratified sampling, a concept that may appear on an exam.
- Considering Specific Topic for Sample Size Search – When seeking previous studies to inform sample size, it is paramount to consider your specific research topic to ensure relevance.
- Sample Size Estimation in Research Proposals – Including detailed sample size estimations in a research proposal is crucial for demonstrating a thorough understanding of sampling and estimation procedures to a review committee.
- Z-value for 95% Confidence Interval – The fixed Z-value of 1.96 for a 95% confidence interval is a value you must know for calculating sample size, particularly for survey-based studies.
- Margin of Error (ME) Value Provision – The specific margin of error value required for sample size calculations (e.g., 5% or 3%) will be provided when needed.
Core Concepts
- Population: The entire group of individuals or units that a researcher is interested in studying; it is also referred to as a parameter.
- Sample: A subset of the population selected by the researcher to participate in a study; it is known as a statistic, and findings from the sample are used to generalise results to the population (inference).
- Sampling Procedure: The systematic process involving defining the target population, establishing a sampling frame that specifies inclusion and exclusion criteria, selecting a sample using an appropriate sampling method, conducting the study on these participants, and then drawing inferences to the wider population.
- Sampling Frame: A comprehensive list or database of all members of the population from which a sample will be drawn, clearly outlining the characteristics required for inclusion in the study. For example, to study cognitive flexibility among individuals with ADHD in Ottawa, the sampling frame would include individuals diagnosed with ADHD by a professional, aged 18 or above, residing in Ottawa, and knowing English.
- Random (Probability) Sampling: A sampling method where every individual in the population has an equal chance of being selected for the study, allowing for greater generalisability of results.
- Simple Random Sampling: Every individual has an equal chance of selection, which can be performed through methods like a lottery system or using computer software (e.g., SPSS). Selection can be ‘with replacement’ (a selected unit can be chosen again) or ‘without replacement’ (a selected unit cannot be chosen again).
- Stratified Sampling: The population is first divided into distinct subgroups (strata) based on shared characteristics (e.g., sex, race), and then random samples are drawn proportionally from each subgroup to ensure representation.
- Cluster Sampling: Entire groups or clusters (e.g., classes, cities) are randomly selected from the population, and then all individuals within the chosen clusters are included in the study. This method is particularly useful for large populations.
- Multi-stage Sampling: A complex sampling technique that combines elements of cluster and stratified sampling. It involves selecting samples in multiple stages, often by first randomly selecting clusters and then randomly selecting individuals from within those chosen clusters.
- Non-random (Non-probability) Sampling: A sampling method where individuals do not have an equal chance of being selected, often relying on the researcher’s judgment or the availability of participants.
- Convenience Sampling: Participants are selected based solely on their availability and ease of access to the researcher.
- Purposive Sampling: Participants are specifically chosen by the researcher based on their expert judgment or because they possess unique characteristics relevant to the study’s objective (e.g., individuals with a specific medical condition).
- Snowball Sampling: Existing study participants are asked to recruit additional participants from their social networks, often used for hard-to-reach populations.
- Quota Sampling: Similar to stratified sampling, this method involves selecting participants to meet predefined quotas for specific subgroups (e.g., ensuring a certain number of males and females) but without the random selection component.
- Volunteer Sampling: A common form of convenience sampling where individuals voluntarily choose to participate in the study based on their willingness and availability, and participation is never mandatory.
- Disadvantages of Non-random Sampling: This approach typically results in more sampling bias (less representative samples), reduced external validity (making it difficult to generalise findings to the broader population), challenges in replication, and weaker statistical bases. However, it is often employed in fields like psychology due to the practical difficulties of accessing entire populations.
- Sample Size Estimation: The process of determining the appropriate number of participants required for a study, crucial for ensuring the reliability and generalisability of results.
- Based on Previous Studies: Involves examining published research papers on similar topics to determine the sample sizes used in comparable studies. The specific design of previous studies (e.g., prevalence, correlational, experimental) should align with the current study’s design. For instance, a previous study on ADHD used 30 participants, suggesting a similar number might be expected.
- Based on Research Type (Rules of Thumb): Provides general guidelines for minimum sample sizes based on the type of statistical analysis planned:
- For correlational studies, a minimum of 50 participants or 10 participants per predictor is suggested.
- For factor analysis, at least 300 participants are recommended.
- For group differences (e.g., using t-tests, ANOVA), a minimum of 30 participants per group is generally advised to achieve 80% power.
- For Chi-square analysis, at least 5 observations per cell are typically required, though 15 per cell is often preferred in practice. A general recommendation for group analyses is at least 15 participants per group.
- Based on Statistical Estimation (Formulas): Utilises mathematical formulas to calculate a precise sample size.
- For correlation/regression, formulas such as Green (n = 50 + 8*m, where ‘m’ is the number of predictors) or Burmeister and Aitken (n = [(p1 + p2 + … + pn) – 1] * 100, where ‘p’ is the number of predictors) can be used. For example, with 5 predictors, Green’s formula suggests 90 participants.
- For cross-sectional or survey studies, a common formula involves the Z-score (1.96 for a 95% confidence interval), the population proportion (often set at 0.5 if unknown), and the margin of error (e.g., 0.05 for 5%). A calculation with these values yields approximately 384 participants.
- Sample size can also be determined by consulting pre-calculated tables that list required sample sizes for various population sizes, given a 95% confidence interval and a 5% margin of error (e.g., for a population of 500, a sample of 217 is suggested).
- Using Statistical Programs: Software like G*Power can assist in computing the expected sample size based on parameters such as margin of error, p-values, confidence intervals, and effect sizes.