Copulas are used to link different types of random variables in statistical models. When dealing with a normal and a half-normal distribution, copulas can help capture their dependence. This allows researchers to create joint distributions with specified marginal distributions, even when the variables are not independent.