Thesis (Ph. D.)--University of Wisconsin--Madison, 1984. / Typescript. Vita. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 143-146).

No description available.

No description available.

Il campionamento a valanga è considerato un tipo di campionamento non probabilistico, la cui rappresentatività può essere valutata solo sulla base di considerazioni soggettive. D’altro canto esso risulta spesso il solo praticamente utilizzabile nel caso di popolazioni senza lista di campionamento. La tesi si divide in due parti. La prima, teorica, descrive alcuni tentativi proposti in letteratura di ricondurre le forme di campionamento a valanga nell’alveo dei campionamenti probabilistici; tra questi è degno di nota il Respondent Driven Sampling, un disegno campionario che dovrebbe combinare il campionamento a valanga con un modello matematico che pesa le unità estratte in modo da compensare la non casualità dell’estrazione e permettere così l’inferenza statistica. La seconda, empirica, indaga le prestazioni del RDS sia attraverso simulazioni sia con una web-survey su una comunità virtuale in Internet, di cui si conoscono la struttura delle relazioni e alcune caratteristiche demografiche per ogni individuo. Le stime RDS, calcolate a partire dai dati delle simulazioni e della web-survey, sono confrontate con i valori veri della popolazione e le potenziali fonti di distorsione (in particolare quelle relative all’assunzione di reclutamento casuale) sono analizzate. / Populations without sampling frame are inherently hard to sample by conventional sampling designs. Often the only practical methods of obtaining the sample involve following social links from some initially identified respondents to add more research participants to the sample. These kinds of link-tracing designs make the sample liable to various forms of bias and make extremely difficult to generalize the results to the population studied. This thesis is divided into two parts. The first part of the thesis describes some attempts to build a statistical theory of link-tracing designs and illustrates, deeply, the Respondent-Driven Sampling, a link-tracing sampling design that should allow researchers to make, in populations without sampling frame, asymptotically unbiased estimates under certain conditions. The second part of the thesis investigates the performance of RDS by simulating sampling from a virtual community on the Internet, which are available in both the network structure of the population and demographic traits for each individual. In addition to simulations, this thesis tests the RDS by making a web-survey of the same population. RDS estimates from simulations and web-survey are compared to true population values and potential sources of bias (in particular those related to the random recruitment assumption) are discussed.

The rapid increase of data volumes makes sampling a crucial component of modern data management systems. Although there is a large body of work on database sampling, the problem of automatically determine the optimal sample for a given query remained (almost) unaddressed. To tackle this problem the authors propose a sample advisor based on a novel cost model. Primarily designed for advising samples of a few queries specified by an expert, the authors additionally propose two extensions of the sample advisor. The first extension enhances the applicability by utilizing recorded workload information and taking memory bounds into account. The second extension increases the effectiveness by merging samples in case of overlapping pieces of sample advice. For both extensions, the authors present exact and heuristic solutions. Within their evaluation, the authors analyze the properties of the cost model and demonstrate the effectiveness and the efficiency of the heuristic solutions with a variety of experiments.