发表机构
University of California, Berkeley; Cornell University(加州大学伯克利分校; 康奈尔大学)
机构由 AI 辅助整理,请以论文原文为准。AI 中文总结
该研究刻画二元逻辑回归对数似然比统计量的有限样本行为,推导不同维度下的非渐近分位数界,其界在目标参数上一致,无需设计正则假设,补充了 Wilks 渐近理论的非渐近结果。
AI 中文摘要
我们刻画了二元逻辑回归中对数似然比统计量的有限样本行为,该行为在设计和目标参数上均一致。对于 $n\geq d\geq 3$,我们在所有固定设计向量集合和所有目标参数上,确定了其最坏情况 $(1-\delta)$ 分位数(相差通用常数倍):$d\log\left(\frac{e n}{d}\right)+\log\left(\frac{1}{\delta}\right)$。这是 Wilks $\chi^2_d$ 现象的非渐近类似物,且无需对设计施加正则性假设。低维情形呈现出异常行为:维度 $d=2$ 时的最坏情况分位数严格为 $\log\log\log n+\log\left(\frac{1}{\delta}\right)$ 量级;维度 $d=1$ 时的最坏情况分位数为 $\log(1/\delta)$ 量级,与 $n$ 无关。最后,独立同分布高斯设计向量可恢复经典 Wilks 尺度。在 $n\gtrsim d+\log(1/\delta)$ 的 regime 下,我们证明了严格界 $d+\log\left(\frac{1}{\delta}\right)$。与现有渐近结果不同,我们的界在目标参数上是一致的,目标参数可依赖于 $n$、$d$ 和 $\delta$。
英文摘要
We characterize the finite sample behavior of the log-likelihood ratio statistic in binary logistic regression, uniformly over both the design and the target parameter. For $n\geq d\geq 3$, we determine, up to universal constants, its worst case $(1-δ)$ quantile over all fixed collections of design vectors and all target parameters: \[ d\log\left(\frac{e n}{d}\right)+\log\left(\frac{1}δ\right). \] This is a nonasymptotic analogue of the Wilks $χ^2_d$ phenomenon and requires no regularity assumptions on the design. The low dimensional cases exhibit unusual behavior. The worst case quantile in dimension $d=2$ is sharply of order \[ \log\log\log n+\log\left(\frac{1}δ\right). \] The worst case quantile in dimension $d=1$ is of order $\log(1/δ)$, with no dependence on $n$. Finally, i.i.d. Gaussian design vectors recover the classical Wilks scale. In the regime $n\gtrsim d+\log(1/δ)$, we prove the sharp bound \[ d+\log\left(\frac{1}δ\right). \] Unlike existing asymptotic results, our bounds are uniform over the target parameter, which may depend on $n$, $d$, and $δ$.
Comments62 pages