arXiv subjects
Publications and source records attributed to Chris Haulk.
Time-varying treatment effects can be estimated with linear regression paired with jackknife confidence intervals. In experiments at a large online video platform, the test based on this estimator is not much less powerful than modern tests for monotonic trends. If you suspect trends in your online experiments, consider using this estimator.
A hierarchy on a set $S$, also called a total partition of $S$, is a collection $\mathcal{H}$ of subsets of $S$ such that $S \in \mathcal{H}$, each singleton subset of $S$ belongs to $\mathcal{H}$, and if $A, B \in \mathcal{H}$ then $A \cap B$ equals either $A$ or $B$ or $\varnothing$. Every exchangeable random hierarchy of positive integers has the same distribution as a random hierarchy $\mathcal{H}$ associated as follows with a random real tree $\mathcal{T}$ equipped with root element $0$ and a random probability distribution $p$ on the Borel subsets of $\mathcal{T}$: given $(\mathcal{T},p)$, let $t_1,t_2, ...$ be independent and identically distributed according to $p$, and let $\mathcal{H}$ comprise all singleton subsets of $\mathbb{N}$, and every subset of the form $\{j: t_j \in F_x\}$ as $x$ ranges over $\mathcal{T}$, where $F_x$ is the fringe subtree of $\mathcal{T}$ rooted at $x$. There is also the alternative characterization: every exchangeable random hierarchy of positive integers has the same distribution as a random hierarchy $\mathcal{H}$ derived as follows from a random hierarchy $\mathscr{H}$ on $[0,1]$ and a family $(U_j)$ of IID uniform [0,1] random variables independent of $\mathscr{H}$: let $\mathcal{H}$ comprise all sets of the form $\{j: U_j \in B\}$ as $B$ ranges over the members of $\mathscr{H}$.
We prove a long-standing conjecture which characterises the Ewens-Pitman two-parameter family of exchangeable random partitions, plus a short list of limit and exceptional cases, by the following property: for each $n = 2,3, >...$, if one of $n$ individuals is chosen uniformly at random, independently of the random partition $π_n$ of these individuals into various types, and all individuals of the same type as the chosen individual are deleted, then for each $r > 0$, given that $r$ individuals remain, these individuals are partitioned according to $π_r'$ for some sequence of random partitions $(π_r')$ that does not depend on $n$. An analogous result characterizes the associated Poisson-Dirichlet family of random discrete distributions by an independence property related to random deletion of a frequency chosen by a size-biased pick. We also survey the regenerative properties of members of the two-parameter family, and settle a question regarding the explicit arrangement of intervals with lengths given by the terms of the Poisson-Dirichlet random sequence into the interval partition induced by the range of a neutral-to-the right process.