An empirical investigation into the impact of winner's curse on estimates from Mendelian randomization