Untitled

• These notes cover research methods including scientific approaches, data collection, research design, ethics, and sampling techniques.

Reasons for Research

• Research reduces bias and subjectivity.
• Research provides systematic and replicable findings.
• Research allows organizations to identify patterns, make informed decisions, and develop evidence-based strategies.

Scientific Approach

• It is systematic, empirical, and based on evidence.
• It follows a structured process of observation, experimentation, and analysis.

Four Major Features of Scientific Process

• Empirical Observation: Based on measurable evidence rather than personal beliefs.
• Logical Reasoning: Uses inductive and deductive logic to form conclusions.
• Skepticism: Findings must be tested and challenged.
• Transparency & Replicability: Other researchers must be able to reproduce results.

Theory and Data

• Theory: An interconnected set of propositions about how or why something occurs.
• Data: Information recorded from observation, which may be numerical or nonnumerical.

Hypothesis

• A testable statement predicts a relationship between variables; for example, employees who receive more feedback have higher job satisfaction.

Verifiable Data

• Data must be observable to the researcher and others in the scientific community, also called "empirical evidence."
• Information needs to be tangible, observed, or sensed through sight, hearing, taste, smell, or touch.

Systematically Collected Data

• It involves the scientific collection and analysis of verifiable data following a typical process or series of steps.
• Skepticism is more likely to yield correct answers and reduces the influence of personal biases.

Empirical Pattern

• This is a recurring trend found in data.
• Example: Remote workers tend to have higher job satisfaction but lower team cohesion.

Inductive vs. Deductive Logic

• Inductive Logic: Starts with specific observations and forms general conclusions.
• Example: Observing that many startups with flexible work hours succeed, concluding that flexible work boosts productivity.
• Inductive logic of inquiry: starts with data, recognizes patterns, and forms conclusions related to theory.
• Deductive Logic: Moves from general principles to specific predictions, progressing from general ideas to specific conclusions.
• Theory states social interaction improves well-being, conducting a study to test if workplace collaboration improves employee satisfaction.

Usefulness of Scientific Approach

• Useful when making high-stakes decisions like hiring policies or marketing strategies and when identifying patterns to predict future trends, such as customer behavior.

Limitations of a Scientific Approach

• Not all human behavior can be quantified.
• Ethical concerns can limit experimentation.

Examples of Questions

• "What is the meaning of life?" exemplifies a philosophical, not empirical, question.

Weak Conclusions

• Inductive reasoning can lead to weaker conclusions because its content extends beyond the provided evidence, making the conclusion uncertain even if the evidence is true.
• Deductive research is precise and allows testing hypotheses.
• Inductive research is exploratory and useful with limited prior studies.

Unit of Analysis

• The level of social life on which a research question focused, or entities (people, nations, artifacts) studied, described, and compared through variables.
• Example: Includes individuals, groups, towns, or nations.

Ecological Fallacy

• Making incorrect conclusions about individual-level processes from group-level data.
• Assuming all employees in a highly productive company are highly motivated.

Cross-Sectional Research Design

• Examines data at a single point in time, using one sample at one moment.
• Surveying employee job satisfaction in 2024 is an example.

Repeated Cross-Sectional Research Design

• Collection of data from different people at multiple time points.
• Conducting employee satisfaction surveys in 2020, 2022, and 2024.

Panel Data Research Design

• Studying the same subjects over an extended period.
• Panel data enables assessing youth's paths into adulthood, lacking historical comparison to preceding generations.

Research Design Comparison

• Cross-sectional research is faster and cheaper, but it is limited in tracking changes.
• Longitudinal research tracks changes over time, but it is more time-consuming and costly.

Qualitative Research

• Deepens understanding through in-depth analysis; asks about social processes or cultural significances.
• The aim is to establish meaning or develop theory.

Quantitative Research

• Uncovers general relationships; examines empirical relationships between two or more variables.
• One aim is to test hypotheses or establish causality and focuses on relationships between variables, often to establish causal relationships.

Variable

• It measures concepts that vary across cases or time.

Independent Variable

• The independent variable is the cause in an experiment.
• Example: The amount of training received.

Dependent Variable

• The dependent variable is the effect being measured.
• Example: Employee performance scores.

Causal Evidence

• Variables must be related (association).
• There should be a direction of influence, such as training preceding increased productivity increases.

Antecedent & Intervening Variables

• Antecedent variables come before the IV, e.g., organizational culture influences training programs.
• Intervening variables come between the IV and DV, e.g., training increases motivation, which improves performance.

Systematic Qualitative Data Collection

• It is collected through structured interviews, focus groups, or ethnography.

Good Research Questions

• Good research questions are clear, focused, and answerable using data.
• They must be testable and based on empirical observation.

Belmont Report Principles

• Respect for Persons: Requires informed consent, acknowledging autonomy and protecting vulnerable populations.
• Beneficence: Minimizes harm, and maximizes the benefits for science, humanity, and research subjects, preventing subject harm.
• Justice: Requires equal distribution of research risks and benefits across subjects.

Concerns for Researchers

• Researchers must consider potential harm (physical, emotional), informed consent, deception, and invasion of privacy.
• Researchers manage harm through informed consent, screening, and providing resources to alleviate harm.
• Respect for privacy involves ensuring anonymity and confidentiality, allowing individuals to decide the extent to which their attitudes, beliefs, and behavior are revealed.

Ethical Issues in Research

• Studies should provide understandable answers to pertinent research questions.
• Invalid research is unethical due to wasted resources and potential risks, leading to poor policy and science.

Unethical Research

• The Rosenhan experiment, where psychologists faked mental illness, was unethical as it risked harm to real patients.

Conceptualization

• Involves defining and clarifying the meaning of a given concept through concise definitions for empirical observations

Operationalization

• This includes identifying the procedures and empirical indicators to measure a concept.

Indicators

• Something that provides evidence or measures a concept’s presence, absence, or degree.
• Many concepts have multiple dimensions, requiring multiple measures.

Level of Measurement

• Assigning numbers or labels to units of analysis.
• The selected measurement level impacts the data analysis.
• You can collapse into lower levels, but cannot disaggregate after collection has been completed.

Measures

• Reliability: Consistency of results under the same conditions.
• Validity: The degree to which results accurately measure what they're supposed to
• The U.S. Census counts every U.S. resident every 10 years, as mandated by the Constitution.

Dangers in Measures

• Dangers of not including enough/the right categories in a measure include:
  • Erasure of populations
  • Undercounting
  • Masking heterogeneity
  • Inadequate policy responses
  • Reinforcing outdated concepts
  • Misallocation of resources

Population vs. Sample

• Population: The entire group of interest.
• Sample: A subset of the population studied.

Random vs. Non-Random Selection

• Random selection ensures equal chances of being chosen.
• Bias may be present in non-random sampling.

Sample Size

• Larger sample sizes reduce error and improves accuracy.

Sampling Techniques

• Simple Random Sample: Every individual has an equal chance of being selected.
• Stratified Random Sample: Split into groups based on characteristics.
• Cluster Sampling: Randomly selecting groups rather than individuals.

Standard Error

• The standard deviation of a statistic's sampling distribution, indicating precision, with larger errors suggesting more variability.
• Each unit of the population should have an equal chance of being selected.

Stratified Random Sample

• Intentionally divides the population into strata.
• This is cheaper and reduces the margin of error.
• Appropriate stratum size is the share of the total population.
• Some groups may be more or less present in a disproportionate stratified sample than they are within the total population.
• Cluster samples allow more surveys at lower cost.

Cluster Sampling

• Cluster sampling allows researchers to take probability samples when no population member list exists.
• Nonprobability samples choose cases other than by random selection when random selection isn't possible or aim is developing holistic understanding via expert judgement.

Purposive Sample

• Purposive samples are guided by expertise on which units best address research questions.
• Snowball samples identify members of a target population, that nominate others.
• This is common mostly among hidden populations with no sampling frame.
• Theoretical sampling iteratively recruits participants to test emerging theories.
• Statistical inference uses data to make conclusions about populations whereas theoretical inference is based on logical reasoning with concepts, and abstract math models.

Sample Cases

• Sampling should continue until little or no new information can be extracted from the data, which is known as saturation.