Suppose that model (7.24) holds for a 2 × J table with J > 2, and let x₂ – x­₁ = 1. Explain why local log odds ratios are typically smaller in absolute value than the cumulative log odds ratio β. [McCullagh and Nelder (1989) noted that local odds ratios {θ_ij} relate to β by logθ_ij = β[P(Y ≤ j + 1) – P(Y ≤ j – 1)] + o(β), j = 1,...,J – 1, where o(β)/β → 0 as β → 0.]