SR TSY 001547 : ISSUE 1

1 Introduction
2 Executive Summary
3 Relationship of This Special Report to Other Bellcore
             Documents
4 General Needs for Proper Analysis
5 Caveats on Procedures Discussed
6 Analysis of Counts of Software Failures
6.1 Models with Only Time As an Explanatory Variable
6.1.1 Testing for Trend
6.1.2 Estimation in the Model with No Trend
6.1.3 Estimating the Magnitude of a Trend
6.1.4 Confidence Interval for an Estimated Trend
6.1.5 Predicting Future Failure Counts with a
             Trend Present
6.2 Performance Comparison of Software Products
6.3 Introduction to Poisson Regression
6.4 Poisson Regression with Only One Explanatory
             Variable
6.5 Poisson Regression with Multiple Explanatory
             Variables
6.5.1 Variable Selection: Judging the Effect of a Variable
6.5.2 Prediction
7 Analysis of Time to Repair Data
7.1 Estimation with Censored Data
7.1.1 Parametric Methods
7.1.2 Nonparametric Methods
7.2 Proportional Hazards Model
7.3 Goodness of Fit Procedures
7.4 Tests for Trends in Repair Times
7.4.1 Testing Differences between Releases
7.4.2 Tests for Repair Time Changes As the Release
             Age Increases
7.4.3 Tests for Abrupt Repair Time Change As the
             Release Age Increases
7.4.4 Other Tests for Trend
7.5 Conclusion
8 Analysis of Other Software Measurements
8.1 Software Changes and Patches
8.2 Problem Reports
REFERENCES
LIST OF FIGURES
Figure A.1 Graph for Maximum Likelihood Estimation of Trend
LIST OF TABLES
Table 6.1 Air Force Dataset
Table 6.2 Maximized Likelihoods for All Possible Models
Table 7.1 Repair Time Data for Two Releases
Table 7.2 Repair Time Data and Fault Detection Time
Table 7.3 Repair Time Data and Fault Detection Time
Table A.1 Approximate Trend Calculations

Bellcore's view of the methods and procedures that should be used in the analysis of software reliability data with these methods, the user can obtain maximum information from software reliability data. This document contains descriptions of some general principles for proper analysis of this data, description of procedures for analyzing counts of software failures and a description of procedures for analyzing time to repair software failure.