A paired test design involves collecting multiple samples from an individual, one corresponding to a control situation and the other to a treatment situation. This allows us to estimate the effect on particular individuals. $ Y_i=X_i^{\text{treatment}} - X_i^{\text{control}} $ ![[paired-test.png|center|400]] We model the difference between each pair of values. A null hypothesis states that the treatment has no effect, which is the same as say that $\mathbb E[Y_i]=0$. In a paired test, the setup of [[Experimental Design]] to randomly allocate between treatment and control is not needed anymore. However, the order of treatment and control trials for each individual need to be randomized, to ensure double-blindness. >[!note:] >Double-blindness means that both the study participant and the researcher do not know if the participant gets the drug or the placebo in a trial.