Randomization does not justify logistic regression

被引：96
作者：
Freedman, David A. [1 ]
机构：
[1] Univ Calif Berkeley, Dept Stat, Berkeley, CA 94720 USA
来源：
STATISTICAL SCIENCE | 2008年 / 23卷 / 02期
关键词：
models; randomization; logistic regression; logit; average predicted probability;
D O I：
10.1214/08-STS262
中图分类号：
O21 [概率论与数理统计]; C8 [统计学];
学科分类号：
020208 ; 070103 ; 0714 ;
摘要：
The logit model is often used to analyze experimental data. However, randomization does not justify the model, so the usual estimators can be inconsistent. A consistent estimator is proposed. Neyman's non-parametric setup is used as a benchmark. In this setup, each subject has two potential if treated and the other if untreated; only one of the two re-responses, can be observed. Beside the mathematics, there are simulation results, a brief review of the literature, and some recommendations for practice.
引用
收藏
页码：237 / 249
页数：13
相关论文
共 49 条
  • [1] AMEMIYA T, 1981, J ECON LIT, V19, P1483
  • [2] Amemiya Takeshi., 1985, Advanced Econometrics
  • [3] [Anonymous], JASA
  • [4] [Anonymous], INT STAT REV
  • [5] [Anonymous], 1935, Supplement to the Journal of the Royal Statistical Society, DOI DOI 10.2307/2983637
  • [6] [Anonymous], 1845, ACAD R SCI METR BR 2
  • [7] ARIS EMD, 2000, P 5 INT C SOC SCI ME
  • [8] Berk RA., 2004, Regression analysis: a constructive critique
  • [9] Application of the logistic function to bio-assay
    Berkson, J
    [J]. JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1944, 39 (227) : 357 - 365
  • [10] Brady HenryE., 2010, Rethinking Social Inquiry: Diverse Tools, Shared Standards, V2nd
12345