Research design, measurement and sampling
A survey begins with a defined research purpose, population, unit of analysis and reference period. Convert an abstract construct into observable indicators without assuming that one convenient question measures every aspect. An operational definition should be clear enough for another investigator to apply consistently. Reliability concerns consistency; validity concerns the interpretation supported by the measurement. A stable but systematically miscalibrated instrument can be reliable without providing the intended valid measurement.
A probability sample uses known selection probabilities, whereas an opt-in sample does not become random merely because many people respond. Distinguish stratification, which samples within groups, from cluster sampling, which selects groups and then observes units within them. A systematic sample needs a random start and attention to periodic patterns in the frame. A missing part of the population is a coverage problem, not random sampling fluctuation.
Choose questions that ask one concept at a time and provide appropriately exclusive and exhaustive response categories. Pilot the wording, sequence and instructions before fieldwork. Random sampling supports population inference under the design; random assignment supports a treatment comparison under appropriate conditions. Neither is a substitute for the other, and an observational association alone does not establish causation.