STATISTICAL TECHNIQUES IN SOFTWARE SIX SIGMA AND DESIGN FOR SIX SIGMA (DFSS) - Software Design for Six-Sigma: A Roadmap for Excellence - page 128

TABLE 6.2 Modeling and Statistical Methods

Statistical Modeling Statistical Methods Modeling Skills

Software metrics input
modeling

- Sampling techniques
- Probability models
- Histograms
- Theoretical distributions
- Parameter estimation
- Goodness-of-fit
- Empirical distributions

- Data collection
- Random generation
- Data classification
- Fitting distributions
- Modeling variability
- Conformance test
- Using actual data

- Graphical tools
- Descriptive statistics
- Inferential statistics
- Experimental design
- Optimization search
- Transfer function
- Scorecard

Software metrics
output analysis

- Output representation
- Results summary
- Drawing inferences
- Design alternatives
- Optimum design

For example, the extracted biohuman material purity from one software to an-
other and the yield of a software varies over multiple collection times. A CTQ
can be cascaded at lower software design levels (system, subsystem, or component)
where measurement is possible and feasible to functional requirements (FRs). At the
software level, the CTQs can be derived from all customer segment wants, needs,
and delights, which are then cascaded to functional requirements, the outputs at the
various hierarchical levels.
Software variables can be quantitative or qualitative. Quantitative variables are
measured numerically in a discrete or a continuous manner, whereas qualitative
variables are measured in a descriptive manner. For example, the memory size of
software is a quantitative variable, wherease the ease of use can be looked at as a
qualitative variable. Variables also are dependent and independent. Variables such as
passed arguments of a called function are independent variables, whereas function-
calculated outcomes are dependent variables. Finally, variables are either continuous
or discrete. A continuous variable is one for which any value is possible within
the limits of the variable ranges. For example, the time spent on developing a DFSS
project (in man-hours) is a continuous variable because it can take real values between
an acceptable minimum and 100%. The variable “Six Sigma Project ID” is a discrete
variable because it only can take countable integer values such as 1, 2, 3 ,etc.It

is clear that statistics computed from continuous variables have many more possible
values than the discrete variables themselves.
The word “statistics” is used in several different senses. In the broadest sense,
“statistics” refers to a range of techniques and procedures for analyzing data,

...