1. Unbiased estimation has been widely considered both in the context of the linear model and more generally, although the limitations of exact unbiasedness as a criterion are widely recognized. The central inequality (10) is essentially due to Fisher (1925), although it was first given in its present form by Cramér (1946) and Rao (1945)
2. see also Aitken and Silverstone (1942).
3. Further bounds involving higher derivatives are due to Bhattacharya (1950)
4. generalizations to nonregular problems are given by Chapman and Robbins (1951) and Kiefer (1952) and to sequential estimation by Wolfowitz (1947).
5. The use of sample splitting to eliminate bias was suggested by Quenouille (1949, 1956) and in the form discussed here by Tukey