Mathematical tools for variables linked by a latent variable
Researchers developed a way to apply Malliavin calculus to random variables that are conditionally independent given a latent random variable .
The work focuses on a countable family of random variables. These variables are conditionally independent given a latent random variable .
Given , the random variables are independent. This allows Malliavin calculus for independent random variables to apply for each value of .
The researchers show these constructions fit together. They obtain the chaos decomposition and the domains of the Malliavin operators. They also find the Mehler formula and the carré du champ operator.
These results use conditional expectations given instead of ordinary expectations. For normal approximation, they use the infinite de Jong theorem of Döbler. This provides mixed normal approximations of degenerate U-statistics in Wasserstein and Kolmogorov distances.
The fluctuations of the conditional variance appear as an additional term. This term cannot be removed.
The study applies these methods to subhypergraph counts in a random -uniform hypergraph. This hypergraph is built on a latent random graph. For the count centered at its conditional mean, they obtain an explicit rate in Wasserstein distance.
Key facts
- The constructions use conditional expectations given instead of ordinary expectations.
- Fluctuations of the conditional variance appear as an additional term that cannot be removed.
- The method provides mixed normal approximations of degenerate U-statistics.
- The work applies to subhypergraph counts in a random -uniform hypergraph built on a latent random graph.