Estimators for Sequential and Simultaneous Nested NPIV
In this section, we analyze the closed-form or approximate solutions under different function classes for the following estimators:
Sequential Nested NPIV:
Given observations \((A_i, B_i, C_i)\) in tr, an initial estimator \(\hat{g}\) which may be estimated in tr, and hyperparameter values \((\lambda, \mu)\), estimate
where \(\text{penalty}(f, \lambda) = \mathbb{E}_m\{f(C)^2\} + \lambda \cdot \|f\|^2_{\mathcal{F}}\) and \(\text{penalty}(h, \mu) = \mu \cdot \|h\|^2_{\mathcal{H}}\).
Sequential Nested NPIV: Ridge:
Given observations \((A_i, B_i, C_i)\) in tr, an initial estimator \(\hat{g}\) which may be estimated in tr, and a hyperparameter \(\mu\), estimate
where \(\text{penalty}(f) = \mathbb{E}_m\{f(C)^2\}\) and \(\text{penalty}(h, \mu) = \mu \cdot \mathbb{E}_m\{h(B)^2\}\).
Simultaneous Nested NPIV:
Given observations \((A_i, B_i, C_i, C_i')\) in tr, and hyperparameter values \((\mu', \mu)\), estimate
using analogous \(\text{penalty}\) notation to the Sequential estimators.