A residual-prediction test with user-chosen ML and sample splitting delivers asymptotic type I error control for linear IV model specification under mean independence of structural errors and instruments, with extensions to weak instruments.
Then, it holds that ∥ ¯A−1 n − A−1 n ∥op = oP(1) (where we set ∥ ¯A−1 n − A−1 n ∥op = ∞ if ¯An is not invertible)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ME 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Machine-Learning-Powered Specification Testing in Linear Instrumental Variable Models
A residual-prediction test with user-chosen ML and sample splitting delivers asymptotic type I error control for linear IV model specification under mean independence of structural errors and instruments, with extensions to weak instruments.