SW Metrics Flashcards
Know metrics for measuring software processes.
What are the different types of scales?
What type of scale is a school grading system?
Nominal. categories: red, green, etc..
Ordinal. comparable set items: bad, good, v/good, etc…
Interval. ordered set with magnitude and distance defined and arbitrary 0 point: calendar dates…
Ratio. like interval except has “absolute” zero point: number of days…
Absolute. count number of items: count LOC…
School grading systems are a ratio scale if they are on a regular-intervaled, numbered system with an absolute 0 (like 0-10, with regular interval 0,5 or 1) because:
- it fulfils the interval scale definition with the following changes and additions:
- it has an absolute 0.
- It also has the ability to say a grade of 2x (8) is twice as good as x (4).
Why is it sometimes good to state some rates in the “per thousand”?
Cause the number can be really low and we want to express it in larger numbers.
Cause sometimes is doesn’t make sense to talk about less than 1 rate of some things that can’t physically be split up (e.g. 0.02 births per person.. rather 22 births per 1000 people).
Is “six sigma” reasonable in software development?
Six sigma, from its original physical manufacturing, is about achieving a defect rate of 3.4 defective parts per million. In software it’d be 3.4 defects per million LOC.
This is unreasonable in software development due to it involving humans, not just machines, in its process.
Rather six sigma involvement in software development is more about the practices behind six sigma. E.g. measurements and statistical methods, process improvement, customer-orientation.
In measurement quality…
What is reliability?
What is validity?
What is the reliability and validity of an I.Q. test?
Reliability: measuring the same thing multiple times gives the same results.
Validity: does the measurement really measure what we intended to measure? (might have to measure things indirectly).
About an I.Q. test: An I.Q. test could be reliable if it always returns the same result when taken by the same person. But as a human is a variable (and perhaps two if including an operator of the test) then it’s likely to be unreliable.
Its validity depends on what we accept as the meaning of ‘intelligence’. If the meaning of intelligence is the result of an I.Q. test (how well someone can solve the things on a specific test) then it defines it itself so will always be valid.
What is GQM?
How many steps are in it.
Goal Question Metrics.
Guide for measuring the right properties using the right metric. Making sure your metrics are useful for answering your questions which are useful for reaching your goal.
Has 3 main steps (determine G and Q, and M). But M is split in 3 steps itself (identify metrics, apply metrics, collect and validate metrics), so that makes 5 steps. Then can add
What are the Test Levels in SW development?
Unit tests: test a single class by calling its methods. Component integration test: test interworking of multiple classes. System test: test whole software via its UI.
What is the “S-curve”?
The shape of the curve formed by cumulative number of tests executed during testing phase.
