Research Design Types

Experimental designs use random assignment to treatment and control groups, ensuring baseline comparability and allowing causal inference. Quasi-experimental designs lack random assignment; instead, they use matching, statistical adjustment, or interrupted time-series to approximate comparability. Quasi-experiments are more practical in real-world settings but are more vulnerable to bias and confounding.

Observational studies can suggest or support causal hypotheses (especially if findings align across multiple studies and mechanisms are plausible), but they cannot definitively establish causality because unmeasured confounders may explain associations. Strong assumptions (e.g., no unmeasured confounding, correct causal model) are required to infer causality from observational data. RCTs remain the gold standard for causal inference.

Use mixed methods when your research question genuinely requires both quantitative and qualitative evidence. For example: 'What is the prevalence of depression in postpartum women, and what are the lived experiences and barriers to care?' Here, prevalence requires surveys or registry data (quantitative); understanding experiences and barriers requires interviews (qualitative). Avoid mixing methods if one approach alone answers your question adequately.

Internal validity is the ability to infer causality within your study (did X cause Y?). External validity is the generalizability of findings to other populations or settings. Both matter, but for different reasons: internal validity allows causal claims; external validity determines whether those claims apply elsewhere. A study strong in internal validity but weak in external validity (e.g., a tightly controlled lab RCT) yields robust causal knowledge about a narrow population. The ideal is balance, though trade-offs are common.