Research Methodology Vocabulary — Experiments, Case Studies & Survey Design
Learn research methodology vocabulary: controlled experiments, case studies, survey design, longitudinal studies, replication studies, sample size, statistical power, and confounding variables.
0 / 5 completed
1 / 5
What is the defining characteristic of a 'controlled experiment' in research methodology vocabulary?
Controlled experiment vocabulary: independent variable (what you manipulate), dependent variable (what you measure), control group (baseline, no treatment), treatment/experimental group (receives the manipulation), random assignment (participants randomly allocated to groups to avoid selection bias). 'Controlled' refers to controlling variables, not the environment. Controlled experiments support causal language: 'X caused Y.' Observational studies only support correlational language: 'X was associated with Y.'
2 / 5
When is a 'case study' methodology appropriate in technical research, and what are its limitations?
Case study language: 'We conducted a case study of a 200-developer organisation migrating from monolith to microservices.' Strengths: rich contextual detail, examines real-world complexity, appropriate for exploratory research. Limitations: 'The findings of this case study may not generalise to organisations of different sizes or industries' (external validity threat). 'Our observations at Company X may reflect its specific culture rather than the technology itself' (confound). Multiple-case studies (comparing several organisations) increase external validity.
3 / 5
What is a 'longitudinal study' and how does it differ from a 'cross-sectional study' in research design vocabulary?
Longitudinal study language: 'We tracked 50 development teams over 18 months, measuring deployment frequency at each quarter.' Advantage: reveals trends, development, and causal sequences. Disadvantage: attrition (participants drop out), expensive, time-consuming. Cross-sectional: 'We surveyed 500 developers at a single point in time.' Advantage: fast and cheap. Disadvantage: cannot distinguish age effects from cohort effects; cannot show causality. A longitudinal study of developer productivity over a career cannot be replaced by a cross-sectional snapshot.
4 / 5
What is a 'confounding variable' and why must it be addressed in research methodology?
Confounding variable example: studying whether code review tools improve code quality. A confound: team experience level. Experienced teams may both use better tools AND produce higher quality code. If you do not control for experience, you may incorrectly attribute the quality improvement to the tool. Language: 'We controlled for team experience by stratifying our sample.' Or: 'Team size was a potential confound — we address this by...' Failing to address confounds: 'A limitation of this study is that we could not fully control for organisational culture, which may confound our results.'
5 / 5
What is 'statistical power' and why is it reported in research methodology sections?
Statistical power language: 'We conducted an a priori power analysis (alpha = 0.05, power = 0.80, Cohen's d = 0.5) which indicated a minimum sample size of 64 per group.' A study with low power (e.g., 0.30) has a 70% chance of failing to detect a real effect — underpowered studies produce unreliable null results. Convention: 0.80 power (80% chance of detecting a true effect). Reporting power shows reviewers the study was not merely under-resourced. Post-hoc power analysis after a null result: 'The achieved power of 0.42 suggests the study may have been underpowered to detect small effects.'