Regression estimation from an individual stable sequence

We consider univariate regression estimation from an individual (non-random) sequence (x₁, y₁), (x₂, y₂),...∈ ℝ x ℝ, which is stable in the sense that for each interval A ⊆ ℝ, (i) the limiting relative frequency of A under x₁, x₂,... is governed by an unknown probability distribution μ, and (ii) the limiting average of those y_i with x_i∈ A is governed by an unknown regression function m(·). A computationally simple scheme for estimating m(·) is exhibited, and is shown to be L₂ consistent for stable sequences {(x_i, y_i)} such that {y_i} is bounded and there is a known upper bound for the variation of m(·) on intervals of the form (-i, i], i ≥ 1. Complementing this positive result, it is shown that there is no consistent estimation scheme for the family of stable sequences whose regression functions have finite variation, even under the restriction that x_i∈ [0, 1] and y_i is binary-valued.