want to construct confidence intervals around those parameters or compute significance tests relating to those parameters, then the assumption of normality matters in small samples, but because of the central limit theorem we don’t really need to worry about this assumption in larger samples
