Software Measurement: A Necessary Scientific Basis

1
Software Measurement A Necessary Scientific Basis

• By Norman Fenton
• Presented by Siv Hilde Houmb
• Friday 1 November

2
Outline

• What is measurements?
• How to measure
• Direct and indirect measurement
• Representation of measurement
• Use of measurement
• Software measurement (different models)

3
What is measurements?

• Measurement is
• The process by which numbers of symbols are
  assigned to attributes of entities in the real
  world in such a way as to describe them according
  to clearly defined rules.
• Entity Object (person, journey, software
  specification)
• Attribute Feature (height, cost, the length of a
  specification)

4
How do we measure?

• What do we mean by assigning values?
• Describe the world
• Need roles on how to evaluate what is the
  longest, biggest etc.
• Models of the entities being measured
• Model must be specific in terms of describing
  phenomena we are studying

5
Direct and indirect measurement

• Direct measurement
• Measurement which does not depend on the
  measurement of any other attribute.
• Indirect measurement
• Measurement which involves the measurement of one
  or more other attributes.

6
Uses of measurement

• Assessment
• Collect data and assess (bedømmelse)
• Prediction
• Model
• Need to define procedure for
• Determine model parameters
• Interpret results
• Prediction system
• Model and procedure for determine model
  parameters and interpreting the result.

7

• Measurement activity must have clear
  objectivities
• Assessment or prediction
• Entity and attribute
• Goal/Question/Metric GQM (Basili and Rombach)

8
Representation of measurement

• Empirical Relation Systems
• Empirical relation system (C,R), where C is the
  set of entities and R is the set of empirical
  relationship.
• Representation Condition
• Mathematically representation of an expression,
  for example taller than (what does that really
  mean (gt) )
• Numerical relation system (N,P), where entities
  in C are mapped to numbers in N and empirical
  relations in R are mapped to numerical
  representation in P.

9
Representation of measurement

• Scale types and Meaningfulness
• Measurement scale (inches, centimetres, scale for
  criticality in risk assessment etc.)
• Not only just a number, but the number relates to
  a scale

10
Measuring Software Complexity

• Comprehensibility
• Correctness
• Maintainability
• Reliability
• Testability
• Ease of implementation

11
Framework for Software Measurement

• Three classes of entities of interest
• Processes
• Products
• Resources
• Internal attributes
• Those which can be measured purely in terms of
  the product, process or resource itself. (eks.
  Length of product)
• External attributes
• Those which can only be measured with respect to
  how the product, process or resource relate to
  other entities in its environment. (eks.
  Reliability)

12
Software metrics

• Cost modelling
• Predicting the attribute of effort or time
  required for the process of development
• Eks. Size number of delivered source statements
  or function points

13
Software quality models

• Quality measured as
• Factors (external product attributes)
• Criteria (internal product/process attribute)
• Metrics (measurement of internal attributes)

14
Reliability models

• Measures external product attributes
• Predictive
• Reliability modelling is concerned with
  predicting reliability of software on the basis
  observing times between failures durig operation
  or testing.

15
Reliability models cont.

• Internal attributes of processes are used to
  predict an external product attribute.
• Prediction system
• Probability distribution model together with a
  statistically inference procedure for determining
  the model parameters, and a prediction procedure
  for combining the model and the parameters
  estimates to make a statement about future
  reliability.

16
Halsteads software science

• Measure
• Length
• Vocabulary
• Volume
• Measure in terms of numbers of operators and
  operands.
• Does not mean program complexity!

17
Validating software measures

• Demonstrating empirically that the representation
  condition is satisfied for the attribute being
  measured
• Needs prediction system and experimental
  hypothesis

18
Summary

• We need to measure internal attributes to support
  the measurement of external attributes (eks.
  Reliability)
• The representation theory of measurements asserts
  that measurements is the process of assigning
  numbers or symbols to attributes of entities in
  such a way that all empirical relations are
  preserved

19
Summary cont.

• The entities of interest in software can be
  classified as processes, products and resources.
• In many cases we cannot measure directly external
  attributes -gt must measure indirectly through
  internal attributes