research-methods

Match the method to your question type: surveys for perceptions at scale, case studies or interviews for why/how questions in context, experiments for causal claims, design science for building and evaluating tools, Delphi for expert consensus, and simulation for systems you cannot field-test. Then apply that method's specific validity checklist.

Choose among convergent (quant and qual in parallel), explanatory sequential (quant then qual to explain), or exploratory sequential (qual to build an instrument, then quant) designs. State the design explicitly, define the integration point, and justify why mixing methods adds value rather than bolting interviews onto a survey.

Use rules of thumb based on the ratio of respondents to indicators, alongside power analysis where possible. Report the measurement and structural models, convergent and discriminant validity, fit indices, and reliability metrics such as Cronbach's alpha and composite reliability.

Use design science research when your contribution is an artifact such as a framework, tool, algorithm, or ontology that must be built and rigorously evaluated. Use a case study when you are investigating a phenomenon within a real project or organization. DSR demands demonstration and evaluation, not just a prototype.

Declare the tool and version, weather data, and model assumptions, then calibrate against measured data and run uncertainty and sensitivity analysis such as Morris or Sobol methods. Distinguish verification (model built right) from validation (right model against empirical data), and report inputs for reproducibility.

A single building or project rarely generalizes, which is AEC research's chronic external validity weakness. Mitigate this with triangulated data sources, a case-study protocol and database, a chain of evidence, thick description for transferability, and explicit acknowledgment of generalization limits.