EasyDeepLearn

What is the delta method?

hard

Answer

  • For a differentiable function g and asymptotically normal θ̂: g(θ̂) ~ N(g(θ),  g(θ)2    Var(θ))N(g({\theta}), \;g({\theta})^{2}\; \cdot \;\operatorname{Var}({\theta})) approximately.
  • Use to derive SE of transformations (ratios, log-odds).
  • Example: Var(log(p))\operatorname{Var}(\operatorname{log}(p))Var(p)  /  p\operatorname{Var}(p)\; / \;p̂².
  • Central for A/B testing ratio metrics (CTR per user, revenue per user with per-user denominator).
  • Fast alternative to bootstrap when you have a closed-form derivative.
Check yourself — multiple choice
  • Random
  • SE(g(θ))\operatorname{SE}(g({\theta})) ≈ |g'(θ)| · SE(θ)\operatorname{SE}({\theta}); used for ratios / log-odds / A/B ratio metrics — closed-form alt to bootstrap when derivative exists
  • Same as bootstrap
  • Only for MLE

Delta method: SE of g(θ̂) via g'(θ); central for ratio metrics.

#estimation#confidence-interval#ab-testing

Practise Statistics Fundamentals

215 interview questions in this topic.

Related questions