Information Technology Reference
In-Depth Information
Check sheet
Pareto Chart
Flowchart
Cause-and-effect diagram
Histogram
Scatterplot
Control chart
FIGURE 6.1
Seven basic quality tools.
techniques are needed when the purpose of the analysis is more complex than just
monitoring the performance of a repeated process element. Regression analysis may
help to optimize the performance of a process.
Development and calibration of effort, quality, and reliability estimation mod-
els often employs regression. Evaluation of alternative processes (e.g., design and
inspection methods) often involves analysis of variance (ANOVA). Empirical soft-
ware research also makes extensive use of ANOVA techniques. The most commonly
employed regression and ANOVA techniques assume that the data under analysis
follows a normal distribution. Dealing with the small samples is common in software
DFSS and that assumption can be problematic. The nonparametric counterparts to
the techniques based on the normal distributions should be used in these situations.
Industry use of statistical techniques is being driven by several standards and ini-
tiatives. The Capability Maturity Model Integration (CMMI) requires the “statistical
management of process elements” to achieve Maturity Level 4 (Emam & Card, 2002).
The latest revisions of ISO Standard 9001 have substantially increased the focus on
the use of statistical methods in quality management.
6.2
COMMON PROBABILITY DISTRIBUTIONS
Table 6.1 is a description of common probability distributions.
6.3
SOFTWARE STATISTICAL METHODS
Statistical methods such as descriptive statistics, removing outliers, fitting data dis-
tributions, and others play an important role in analyzing software historical and
developmental data.
The largest value added from statistical modeling is achieved by analyzing soft-
ware metrics to draw statistical inferences and by optimizing the model parame-
ters through experimental design and optimization. Statistics provide a flexible and
Search WWH ::




Custom Search