자료유형  

 학위논문

Control Number  

0016934856

International Standard Book Number  

9798380333542

Dewey Decimal Classification Number  

004

Main Entry-Personal Name  

Jun, Eunice.

Publication, Distribution, etc. (Imprint  

[S.l.] : University of Washington., 2023

Publication, Distribution, etc. (Imprint  

Ann Arbor : ProQuest Dissertations & Theses, 2023

Physical Description  

1 online resource(244 p.)

General Note  

Source: Dissertations Abstracts International, Volume: 85-03, Section: B.

General Note  

Advisor: Heer, Jeffrey;Just, Rene.

Dissertation Note  

Thesis (Ph.D.)--University of Washington, 2023.

Restrictions on Access Note  

This item must not be sold to any third party vendors.

Summary, Etc.  

요약Data analysis is critical to science, public policy, and business. Despite their importance, statistical analyses are difficult to author, especially for researchers with expertise outside of statistics. Existing statistical tools, prioritizing mathematical expressivity and computational control, are low-level while researchers' motivating questions and hypotheses are high-level. The process of translating researchers' questions and hypotheses into low-level statistical code is error-prone.This thesis views statistical analysis authoring as a sensemaking process that involves grappling with domain knowledge, statistics, and programming concerns. To this end, I develop a framework characterizing the cognitive and operational steps involved in translating research questions into statistical analysis code, a process I term hypothesis formalization. I also design, implement, and evaluate three new domain-specific languages (DSLs) and runtimes that embody hypothesis formalization. The DSLs leverage automated reasoning to compile high-level specifications of analysis intent into analysis code.The first of these, Tea, is used to author Null Hypothesis Significance Tests. Analysts specify their study design, assumptions about data, and hypotheses in Tea's DSL. Tea represents statistical test selection as constraint satisfaction, so it compiles an analyst's specification into a system of constraints to identify a set of valid statistical tests. A benchmark comparison found that Tea's test selection is comparable to that of experts and better than a naive test selection approach.I also introduce Tisane, a system for authoring generalized linear models with or without mixed effects. Analysts specify their domain knowledge in the form of a conceptual model, data collection details, and focus of analysis in Tisane's DSL. Internally, Tisane represents this conceptual model as a graph. Tisane traverses the graph to derive a space of statistical models based on causal reasoning recommendations. Then, in an interactive disambiguation process, Tisane involves analysts in narrowing the space of possible statistical models to one final output statistical modeling script. In case studies, we found that Tisane shifted researchers' focus from analysis details to their research questions and streamlined the analysis authoring process. To further improve the usability of the Tisane DSL, I conducted an exploratory elicitation study using Tisane as a probe, designed and implemented an improved version of Tisane as rTisane, and then evaluated rTisane in a controlled lab study. The summative evaluation demonstrated that rTisane's DSL helped analysts introspect on their implicit domain assumptions more deeply, stay true to their analysis intent, and produce statistical models that better fit the data. In all, these systems and evaluations provide evidence that conceptually focused DSLs coupled with automated reasoning can lower the barriers to valid analyses.

Subject Added Entry-Topical Term  

Computer science.

Subject Added Entry-Topical Term  

Statistics.

Subject Added Entry-Topical Term  

Cognitive psychology.

Subject Added Entry-Topical Term  

Computer engineering.

Index Term-Uncontrolled  

Data analysis

Index Term-Uncontrolled  

End-user software engineering

Index Term-Uncontrolled  

Human-computer interaction

Index Term-Uncontrolled  

Programming languages

Index Term-Uncontrolled  

Sensemaking

Index Term-Uncontrolled  

Statistical software

Added Entry-Corporate Name  

University of Washington Computer Science and Engineering

Host Item Entry  

Dissertations Abstracts International. 85-03B.

Host Item Entry  

Dissertation Abstract International

Electronic Location and Access  

로그인을 한후 보실 수 있는 자료입니다.

Control Number  

joongbu:643018