Research Design Issues in Health Sciences - Week 3 Part 1

Representative sample

A sample that shares the exact characteristics of the target population for valid generalisation.

Sampling frame

A complete list of the entire population of interest used to draw probability samples.

Convenience sampling

A non-probability method where researchers select participants from whoever is easily available.

Purposive sampling

Deliberately gathering a sample of individuals who all share a specific characteristic.

Reliability

The consistency and reproducibility of a research test or finding over time.

Test-retest reliability

Getting similar outcomes when administering the same measure to the same person repeatedly.

Internal consistency

The degree to which responses within a single measure tell the same story.

Inter-rater reliability

The degree of agreement and consistency between different observers or scorers.

Construct validity

The degree to which a test successfully measures the specific theoretical variable it intends to.

External validity

The extent to which research findings can be generalised to other times, places, or populations.

Internal validity

The confidence that the manipulated independent variable directly caused the observed outcome.

Ecological validity

The extent to which study findings successfully apply to real-world settings.

Demand characteristics

Cues in an experiment that signal to participants how they are expected to respond.

Hawthorne effect

Changes in subject behaviour caused simply by the knowledge that they are being observed.

Experimenter bias

Unconscious researcher behaviour that prevents a fair and objective test of the hypothesis.

Experiment

Participants are randomly assigned to IV conditions and the effect on a DV is measured.

Causality in experiments

Experiments can infer causality if they are conducted appropriately.

Manipulating variables

Experiments involve manipulating Independent Variables and measuring Dependent Variables.

Split-ballot techniques

A survey in which there is manipulation, e.g., different participants receive different questionnaires.

Quasi-experiments

Manipulation without randomisation to treatment groups.

Cross-sectional studies

A sample is gathered across the population of interest at one point in time; possible problems are confounds and identifying causation.

Longitudinal studies

Involve tracking a sample of participants over time; possible problems include mortality, maturation and testing effects.

Successive cross-sectional studies

Involve surveying different samples at different points in time; possible problems include requiring more participants.

Main types of surveys

Questionnaires, interviews, and computer-aided interviewing.

Survey method suitability

No single method is universally better; suitability depends on the specific situation.

Correlation and Causation

Correlation does not necessarily mean causation.

Absence of Relationship

If no relationship exists between variables, we can conclude one does not cause the other (barring hidden third variables).

Surveys vs. Experiments Focus

Surveys tend to be more concerned with description than explanation.

Correlation in Surveys

Correlation does not equal causation; a correlation between two variables does not mean one causes the other.

Naturally-occurring relationships

Measuring variables already out there in the population without changing anything.

Enduring relationships

Associations between different variables that have some stability over time.

Relationships between variables in surveys

Examining whether changes in one variable correspond to changes in another.

Survey vs. Experiment Location

Surveys can be conducted anywhere, while some experiments occur outside the lab and some surveys inside the lab.

Focus of Experiments

Establishing causal relationships between variables by randomly assigning participants to different levels of the IV.

Focus of Surveys

Measuring naturally-occurring and enduring relationships between variables.