🐛Biostatistics Unit 4 – Sampling and Design in Biological Research

Key Concepts

• Sampling involves selecting a subset of individuals from a population to estimate characteristics of the entire population
• Sampling methods can be classified as probability sampling or non-probability sampling
  • Probability sampling uses random selection and gives each member of the population a known chance of being selected
  • Non-probability sampling does not involve random selection and may be subject to bias
• Experimental design is the process of planning a study to meet specified objectives, including selecting treatments, experimental units, and measurement procedures
• Randomization assigns subjects to different treatments randomly to minimize bias and ensure treatment groups are comparable
• Replication involves repeating the basic experiment on multiple subjects to provide an estimate of experimental error and increase precision
• Blocking is a technique used to reduce known variability among experimental units by grouping similar units together (blocks) and assigning treatments within each block
• Sample size determination is critical to ensure the study has sufficient power to detect meaningful differences or effects
• Power analysis estimates the minimum sample size required to detect an effect of a given size with a specified degree of confidence

Types of Sampling Methods

• Simple random sampling selects a subset of individuals from a population such that each individual has an equal probability of being chosen
• Stratified sampling divides the population into subgroups (strata) based on specific characteristics and then performs random sampling within each stratum
  • Ensures representation of key subgroups within the sample
  • Improves precision of estimates for each stratum
• Cluster sampling involves dividing the population into clusters, randomly selecting a subset of clusters, and including all individuals within the selected clusters in the sample
• Systematic sampling selects individuals from a population at regular intervals (e.g., every 10th person on a list)
• Convenience sampling selects individuals who are easily accessible or willing to participate, but may not be representative of the entire population
• Purposive sampling selects individuals based on specific characteristics or criteria relevant to the research question
• Snowball sampling relies on initial subjects to recruit additional subjects among their acquaintances, often used when studying hard-to-reach populations

Experimental Design Basics

• Completely randomized design assigns treatments to experimental units completely at random, ensuring each unit has an equal chance of receiving any treatment
• Randomized block design divides experimental units into homogeneous blocks and randomly assigns treatments within each block to reduce variability and increase precision
• Factorial design investigates the effects of two or more factors simultaneously by testing all possible combinations of factor levels
  • Allows examination of main effects and interactions between factors
  • Provides a more comprehensive understanding of the system under study
• Split-plot design involves applying one set of treatments to large experimental units (whole plots) and another set of treatments to smaller experimental units (subplots) within each whole plot
• Crossover design exposes each subject to a sequence of treatments over time, with each subject serving as their own control
  • Reduces inter-subject variability and requires fewer subjects
  • Suitable when treatment effects are reversible and do not carry over to subsequent periods
• Latin square design is used when there are two sources of variability (rows and columns) in addition to the treatment effect, ensuring each treatment appears exactly once in each row and column

Sample Size and Power Analysis

• Sample size is the number of individuals or experimental units included in a study
• Adequate sample size is crucial for detecting meaningful differences or effects and ensuring the reliability and validity of study results
• Factors influencing sample size include the variability of the response variable, the magnitude of the effect of interest, the desired level of significance, and the power of the test
• Power is the probability of rejecting the null hypothesis when it is false (i.e., detecting a true effect)
  • Depends on sample size, effect size, significance level, and variability of the response variable
  • Increasing sample size, effect size, or significance level increases power
• A priori power analysis determines the minimum sample size required to achieve a desired level of power for a specified effect size and significance level
• Post hoc power analysis estimates the power of a completed study based on the observed effect size and sample size
• Balancing statistical power, feasibility, and ethical considerations is essential when determining appropriate sample sizes for biological research

Bias and Confounding Factors

• Bias is a systematic error that leads to an incorrect estimate of the association between an exposure and an outcome
• Selection bias occurs when the study sample is not representative of the target population due to the way subjects are selected or participate in the study
  • Can be minimized by using random sampling methods and ensuring high participation rates
• Information bias arises from errors in measuring or classifying exposures, outcomes, or other variables of interest
  • Examples include recall bias, interviewer bias, and misclassification bias
  • Can be reduced by using standardized and validated measurement tools, blinding, and quality control procedures
• Confounding occurs when a third variable is associated with both the exposure and the outcome, distorting their true relationship
  • Age, sex, and socioeconomic status are common confounding factors in biological research
  • Confounding can be controlled through study design (randomization, matching) or statistical analysis (stratification, multivariate modeling)
• Randomization helps distribute potential confounding factors evenly across treatment groups, minimizing their impact on the observed treatment effect
• Blinding of participants, investigators, and outcome assessors can help reduce bias by preventing knowledge of treatment assignment from influencing behavior or measurements

Data Collection Techniques

• Surveys and questionnaires are used to gather information directly from participants, often assessing exposures, outcomes, or other variables of interest
  • Can be administered in person, by phone, mail, or online
  • Proper design and validation are crucial for ensuring data quality and minimizing bias
• Interviews involve direct communication between the researcher and the participant, allowing for more in-depth exploration of topics
  • Structured interviews follow a predefined set of questions, while unstructured interviews allow for more flexibility and probing
  • Training of interviewers is important for maintaining consistency and reducing interviewer bias
• Observational methods involve collecting data through direct observation of participants, environments, or processes
  • Can be used to assess behaviors, interactions, or other phenomena that may be difficult to capture through other means
  • Standardized protocols and observer training are essential for ensuring data reliability
• Physical measurements and biospecimen collection provide objective data on physiological, biochemical, or genetic characteristics
  • Standardized procedures, quality control, and proper storage and handling of specimens are critical for ensuring data integrity
• Administrative databases and electronic health records can serve as valuable secondary data sources, providing information on exposures, outcomes, and potential confounders
  • Data quality, completeness, and validity should be carefully assessed when using these sources

Statistical Considerations

• Descriptive statistics summarize and describe the main features of a dataset, such as measures of central tendency (mean, median) and dispersion (standard deviation, range)
• Inferential statistics involve using sample data to make generalizations about the population from which the sample was drawn
  • Hypothesis testing is a common inferential approach, testing the compatibility of the observed data with a null hypothesis
  • Confidence intervals provide a range of plausible values for a population parameter based on the sample data
• Parametric tests assume that the data follow a specific probability distribution (often normal) and have certain properties, such as equal variances across groups
  • Examples include t-tests, ANOVA, and linear regression
  • More powerful than non-parametric tests when assumptions are met
• Non-parametric tests do not rely on assumptions about the distribution of the data and are suitable when assumptions of parametric tests are violated or the data are ordinal or nominal
  • Examples include Mann-Whitney U test, Kruskal-Wallis test, and chi-square test
  • Less powerful than parametric tests but more robust to violations of assumptions
• Multiple testing correction is necessary when conducting multiple hypothesis tests simultaneously to control the family-wise error rate or false discovery rate
  • Bonferroni correction and Benjamini-Hochberg procedure are common methods for adjusting p-values
• Missing data can introduce bias and reduce statistical power if not handled appropriately
  • Strategies for dealing with missing data include complete case analysis, single imputation, and multiple imputation

Real-World Applications

• Clinical trials rely on rigorous experimental design, randomization, and blinding to evaluate the safety and efficacy of new medical interventions (drugs, devices, or procedures)
  • Proper sample size determination and power analysis are crucial for ensuring the trial can detect clinically meaningful effects
  • Stratified randomization and blocking are often used to balance treatment groups with respect to important baseline characteristics
• Epidemiological studies investigate the distribution and determinants of health-related states or events in specified populations
  • Cohort studies follow a group of individuals over time to assess the association between exposures and outcomes
  • Case-control studies compare exposures between individuals with a specific outcome (cases) and those without the outcome (controls)
  • Sampling methods, such as stratified or cluster sampling, are often used to ensure representativeness and efficiency
• Ecological studies examine associations between exposures and outcomes at the population or group level, rather than the individual level
  • Can be useful for generating hypotheses but are prone to ecological fallacy (inferring individual-level associations from group-level data)
  • Careful consideration of potential confounding factors and data quality is essential
• Meta-analysis combines results from multiple studies addressing the same research question to provide a more precise and comprehensive estimate of the effect of interest
  • Requires thorough literature search, data extraction, and assessment of study quality and heterogeneity
  • Helps to synthesize evidence and inform evidence-based decision making in various fields of biological research