Accounting for selection bias due to death in estimating the effect of wealth shock on cognition for the Health and Retirement Study