Experimental & quasi-experimental design

Want to prove your new therapy actually causes improvement, not just that improvement happened alongside it? Then you need a true experiment — and the one ingredient examiners test above all is random assignment.

Random assignment means every participant has an equal chance of landing in the treatment or control condition. That one move is the defining feature of a true experimental design. Why it matters: randomization spreads pre-existing differences (motivation, severity, age) roughly evenly across groups, so any post-treatment difference can be credited to the independent variable rather than to who happened to be in which group. Don't confuse it with random selection (how you sample from a population — that's about generalizing out); random assignment is how you sort your sample into groups.

The control group is the comparison. It gets no treatment, a placebo, or treatment-as-usual, giving you a baseline for what would have happened anyway. Treatment minus control = the effect. The researcher manipulates the independent variable and measures the dependent variable.

All of this serves internal validity: the confidence that the IV, and not something else, produced the change. Random assignment plus a control group is the strongest guard against confounds. (The named catalog of validity threats — history, maturation, testing, and so on — is a separate sibling topic; here just hold the logic that experiments exist to protect causal inference.)

A quasi-experiment looks like an experiment but is missing random assignment — usually because you're stuck with intact, pre-existing groups (two classrooms, two clinics). You still manipulate an IV and use a comparison group, but because assignment wasn't random, the groups may have started out different, weakening causal claims. Quasi-experiments are common and useful in real-world settings where randomizing people is impractical or unethical.

Clinical Pearl: The single line examiners test — random assignment is what makes a design a true experiment; remove it and you have a quasi-experiment. Random assignment = equal chance of group placement (internal validity); random selection = how you draw the sample (generalizability).