FREE Research practice LESSON · Research practice
Measure without fooling yourself
Construct validity, uncertainty, causality, and robustness
A measurement is an argument connecting an observation to a claim.
Construct validity asks whether the metric represents the intended concept; internal validity asks whether the design supports the causal explanation; external validity asks where results transfer. Sampling, confounding, researcher choices, multiple comparisons, and noisy instruments can weaken each link. Robust work makes the estimand, uncertainty, and plausible alternative explanations explicit.
A precise number can still answer the wrong question with great confidence.
Statistical significance does not establish practical importance.
A tiny effect can become statistically detectable with enough observations, while a consequential effect can remain uncertain in a small sample. Report effect sizes, intervals, decision thresholds, and sensitivity to assumptions. When many analyses were possible, disclose selection and protect confirmation with new data.
Ask what decision changes across the credible range of the effect.