Constructing Confidence Intervals for Sensitivity Under Controlled Specificity in Medical Tests Público

Abstract

Abstract

Constructing Confidence Intervals for Sensitivity Under Controlled Specificity in Medical Tests

By: Isaac Parakati

Introduction: Medical tests frequently assist health care workers with identifying individuals affected or not affected by a disease. Although medical tests are supposed to correctly identify diseased individuals as diseased and non-diseased individuals as non-diseased, this does not always occur. The test's accuracy is typically measured in terms of sensitivity and specificity. Fixing the specificity of a test at a particular value, the test's corresponding sensitivity can be determined. The goal of this paper is to propose two new approaches for constructing confidence intervals for sensitivity after fixing specificity.

Methods: To estimate sensitivity, both of the two proposed approaches are based on a quadratic inference function but differ by the procedure used to profile out a nuisance parameter. The first approach minimizes the function with respect to the nuisance parameter. The second approach determines an optimal weighted average between two values for the nuisance parameter. To demonstrate the two approaches, confidence intervals were constructed for the sensitivity of a gene expression biomarker using samples of cancerous and non-cancerous tissues, fixing specificity. Simulations were conducted to evaluate the approaches under different distributions with varying sample sizes. Coverage probabilities and average confidence interval length were determined for each simulation.

Results: In the simulations, the two new approaches produced confidence intervals above or near the nominal significance level. The first approach constructed very wide intervals with conservative coverage. The second approach constructed narrower intervals with coverage near the nominal value; this approach performed similarly to the leading existing BTII approach, whose simulation results were extracted from Zhou and Qin's paper⁴. The BTII approach seemed to perform slightly better when the diseased and non-diseased sample sizes differed. With larger sample sizes, average confidence interval length for all approaches narrowed.

Discussion: The second approach proposed in this paper appears to be a suitable non-bootstrap alternative to the BTII approach when constructing confidence intervals.

Table of Contents

Table of Contents

1. Introduction......................................................................................................................................... 1

1.1 Problem Statement and Notation........................................................................................................... 1

1.2 Purpose Statement.............................................................................................................................. 2

1.3 Significance Statement......................................................................................................................... 2

2. Background/Literature Review................................................................................................................. 2

2.1 Naive Interval..................................................................................................................................... 2

2.2 Linnet Interval.................................................................................................................................... 3

2.3 Bootstrap Intervals.............................................................................................................................. 4

3. Methods............................................................................................................................................... 6

4. Practical Application to Cancer Tissue....................................................................................................... 9

5. Simulations........................................................................................................................................... 9

5. Discussion............................................................................................................................................ 11

6. References........................................................................................................................................... 13

7. Appendix.............................................................................................................................................. 14

7.1 Bisection Method................................................................................................................................. 14

7.2 Tables & Figures.................................................................................................................................. 15

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics and Bioinformatics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	MSPH
Submission	Master's Thesis
Language	English
Research Field	Biology, Biostatistics Statistics Health Sciences, Public Health
Palabra Clave	Medical Tests Diagnostic Tests Confidence Intervals Sensitivity Specificity
Committee Chair / Thesis Advisor	Huang, Eugene, Emory University
Committee Members	Peng, Limin, Emory University
Partnering Agencies	Does not apply (no collaborating organization)

Última modificación

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Constructing Confidence Intervals for Sensitivity Under Controlled Specificity in Medical Tests ()	2018-08-28 15:53:46 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions