12. Which of the following is a reason to prefer weighted least-squares regression (WLS) over strati cation on binned data with continuous Z? (a) If we're missing data in some X;Z strata, there is no way to improve the data density over strata in the strati cation method. (b) Binning is "lossy", so adjustment with binned data will not completely block back-door paths. (c) Strati cation is non-parametric, so is more resilient to model mis specification. (d) IPW allows more flexible model specifications