Cmpe 589

1
Cmpe 589

• Spring 2006

2
Measurement Theory

• Front-End
• Design
• Design Review and Inspection
• Code
• Code Inspections
• Debug and Develop Test Cases
•  
• ? Integration

3
Measurement Theory

• Back-End
• Formal Machine Testing
• Early Customer Programs
•  
• Rigorous Implementation
• Assuming process formally documented
• Adherence to process
•  
• Data Collection and Testing Hypothesis

4
Measurement Theory

• Measurement Scales
• Nominal- Category names, (no ordering),
  categories are mutually exclusive, jointly
  exhaustive
• Ordinal- Categories are ordered
• If A gt B then B is not equal to A
• If A gt B and B gt C then A gt C
•  
• Completely Satisfied
• Satisfied
• Neutral
• Somewhat Dissatisfied
• Completely Dissatisfied

5
Measurement Theory

• Interval
• Constant distance between measurements (unit)
• Arithmetic
• 16 degrees Celsius vs. 80 degrees Celsius
• Ratio- Absolute zero, interval scale

6
Basic Measurements

• Ratio- fraction
• Numerator and denominator are mutually exclusive
  categories (number of males/ number of females)
• Proportion- Numerator of fraction is part of the
  denominator
•   N a b c
• P (a/(abc))
• 1 (a/n) (b/n) (c/n)
• Percentage- Proportion scaled to 100 percent of
  the denominator
• Total Population of 200 defects 15 were in
  requirements, 25 were coding errors, 50 were
  design errors, 10 were from other sources.
• Rate- Dynamic Changes
• Formula in notes
• Bell Curve

7
Basic Measurements

• Validity- the degree to which a value actually
  measures what youre looking at.
• Reliability- Consistency of measurements on some
  individual using same places over time
•  Smaller the index the more reliable
• Index of Variability (I.V.) Standard Deviation
  / mean
•  

8
Basic Measurements

• Validity- Construct validity- operation
  definition matches incorruptible concepts
• Expert panel
• Criterion Validity- degree to which the measure
  covers range meanings of concept
• Expert panel
• Content Validity- degree to which the measure
  covers range meanings of concept
• Expert panel

9
Basic Measurements

• Measurement Errors
• Systematic Error- Validity
• Random Error- Reliability
• General Case
• M T S E
• M Observed Measure
• T Actual Measure
• S Systematic Error
• E random error
• M T E
• Var(M) var (T) var (E)
• Reliability var (T) /var (M)
• var(M) var(E)/ var(M)
• 1 var(E)/var(M)

10
Basic Measurements

• How to Assess Reliability
• Test/Retest
• Alternative Forms
• Split Halves
• Internal Consistency

11
Basic Measurements

• Correlation Cautions
• Correlation expects linear relationship
• Noisy data for outliers affect r or P
  adversely--(Spearmans r) non parametric
  version
• Large data range
• Correlation is not equal to causality

12
Basic Measurements

• Criteria for Causality
• Cause precedes the effect (logic or time)
• Two variables must be empirically or
  experimentally correlated
• Third the relationship is not spurious
  relationship