{"id":"arxiv/2610.07190","url":"https://arxiv.org/abs/2610.07190","tags":["stat.ME","math.ST","stat.ML"],"score":0,"title":"Poisson empirical Bayes estimation of sums of random variables via minimum-distance methods","domain":"stat.ME","source":"arxiv","authors":["Stefano Favaro","Sandra Fortini","Soham Jana"],"reviews":0,"subject":"stat.ME","summary":"The estimation of sums of functions of observable and unobservable variables is a long-standing problem in statistics, with applications in many domains. We consider this problem in Poisson mixture models, where empirical Bayes provides a natural framework but nonparametric theory remains limited. We develop a nonparametric empirical Bayes methodology based on regular and coarsened minimum-distance estimation of the unknown mixing distribution. For a broad class of such sums, we establish large-sample guarantees showing that the resulting plug-in estimates asymptotically merge with the…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-05","reactions":{}}