STATISTICAL TECHNIQUES IN SOFTWARE SIX SIGMA AND DESIGN FOR SIX SIGMA (DFSS) - Software Design for Six-Sigma: A Roadmap for Excellence - page 123

most commonly employed techniques. These techniques involve the rigorous collec-
tion of data, development of statistical models describing that data, and application
of those models to decision making by the software DFSS team. The result is better
decisions with a known level of confidence.
Statistics is the science of data. It involves collecting, classifying, summarizing,
organizing, analyzing, and interpreting data. The purpose is to extract information to
aid decision making. Statistical methods can be categorized as descriptive or infer-
ential . Descriptive statistics involves collecting, presenting, and characterizing data.
The purpose is to describe the data graphically and numerically. Inferential statis-
tics involves estimation and hypothesis testing to make decisions about population
parameters. The statistical analysis presented here is applicable to all analytical data
that involve counting or multiple measurements.
Common applications of statistics in software DFSS include developing effort
and quality estimation models, stabilizing and optimizing process performance, and
evaluating alternative development and testing methods. None of the techniques can
be covered in sufficient detail to develop real skills in their use. 3 However, the chapter
will help the practitioner to select appropriate techniques for further exploration and
to understand better the results of researchers in relevant areas.
This chapter addresses basic measurement and statistical concepts. The approach
presented is based on ISO/IEC Standard 15939 (Emam & Card, 2002). An effective
measurement and analysis program in measurement topics include measurement
scales, decision criteria, and the measurement process model provided in ISO/IEC
Standard 15939. Statistical topics include descriptive statistics, common distributions,
hypothesis testing, experiment design, and selection of techniques. Measurement and
statistics are aids to decision making. The software DFSS team makes decisions on
a daily basis with factual and systematic support. These techniques help to improve
the quality of decision making. Moreover, they make it possible to estimate the
uncertainty associated with a decision.
Many nonstatistical quantitative techniques help to select the appropriate statistical
technique to apply to a given set of data, as well as to investigate the root causes of
anomalies detected through data analysis. Root cause analysis as known today relies
on seven basic tools that are the cause-and-effect diagram, check sheet, control chart
(special cause vs. common cause), flowchart, histogram, Pareto chart, and scatterplot.
They are captured in Figure 6.1. Other tools include check sheets (or contingency
tables), Pareto charts, histograms, run charts, and scattergrams. Ishikawa's practical
handbook discusses many of these.
Although many elements of the software DFSS only are implemented once or a few
times in the typical project, some activities (e.g., inspections) are repeated frequently
in the Verify & Validate phase. Monitoring these repeated process elements can help
to stabilize the overall process elements. Many different control charts are available.
The choice of techniques depends on the nature and organization of the data. Few
basic statistics texts cover control charts or the more general topic of statistical
process control, despite their widespread applicability in industry. Other statistical
3 Contact www.SixSigmaPI.com for training.