# How can I simulate data for comparing logit and probit model?

Can anybody tell me how can I simulate a data set for comparing logit and probit model? If I simulate a data from a standard multivariate normal distribution considering 5 variables then can I use one of them as continuous latent variable (for which I consider non-zero covariance in the covariance matrix)? If I can, then what should be the cut point (of latent variable $y^{*}$) for considering $Y=1$ and $Y=0$. And I need some reference where this short of works are conducted.