Fixed-Size Active Statistical Inference

Key points

• Variance reduction techniques improve the efficiency of drawing conclusions from data.1
• Recent proposals for variance-reduction use two complementary techniques to reduce variance: incorporating machine learning predictions into estimators and weighted sampling.1
• Notably, we show that the asymptotic variance of CPS is no worse than that of PS, and can be substantially better in realistic scenarios where predictions are imperfect.1
• We illustrate these advantages on both synthetic and real datasets, and release a Python package implementing CPS.1

Sources (1)

• [1]Fixed-Size Active Statistical Inference
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 11:25 PM

  We show that a fixed-size alternative, known in survey sampling as Conditional Poisson Sampling (CPS), is a drop-in replacement for PS on general multivariate Z-estimation problems, which include convex M-estimation considered in prior work.

  Variance reduction techniques improve the efficiency of drawing conclusions from data.

Extractive summary: sentences quoted from the sources.