jackson bi thinks this is interesting:

a stock outlier analysis where we regress the model error on stock specific characteristics (such as market cap, spread, idiosyncratic risk, etc.).