返回
Statistical testing under distributional shifts
DOI:10.1093/jrsssb/qkad018.png)
摘要
En 中文
We introduce statistical testing under distributional shifts. We are interested in the hypothesis P*? H(0 )for a target distribution P*, but observe data from a different distribution Q*. We assume that P* is related to Q* through a known shift t and formally introduce hypothesis testing in this setting. We propose a general testing procedure that first resamples from the observed data to construct an auxiliary data set (similarly to sampling importance resampling) and then applies an existing test in the target domain. We prove that if the size of the resample is of order o(vn) and the resampling weights are well behaved, this procedure inherits the pointwise asymptotic level and power from the target test. If the map t is estimated from data, we maintain the above guarantees under mild conditions on the estimation. Our results extend to finite sample level, uniform asymptotic level, a different resampling scheme, and statistical inference different from testing. Testing under distributional shifts allows us to tackle a diverse set of problems. We argue that it may prove useful in contextual bandit problems and covariate shift, show how it reduces conditional to unconditional independence testing and provide example applications in causal inference.
Keyword:
distributional shifts
hypothesis testing
conditional independence testing
sampling importance resampling
causality
期刊
J
IF:
3.6
论文数:
1.5K
被引数:
3.2W
机构
引用论文
Effects of fasting during Ramadan on seizure control and quality of life in patients with epilepsy斋月期间禁食对癫痫患者癫痫发作控制和生活质量的影响
Crystal chemistry and metal-hydrogen bonding in anisotropic and interstitial hydrides of intermetallics of rare earth (R) and transition metals (T), RT3 and R2T7稀土 (R) 和过渡金属 (T) 的金属间化合物的各向异性和间隙氢化物中的晶体化学和金属氢键,RT3 和R2T7

