Software Measurement: Uncertainty and Causal Modeling

1. Software Measurement: Uncertainty and Causal Modeling Koosha Golmohammadi www.kooshanet.com golmoham@ualberta.ca

2. Software Measurement: Uncertainty and Causal Modeling P 1/12 Outlines • What is software • What is software measurement • What is uncertainty • What is software quality • What is risk management • What is Bayesian method • What is causal modeling and Bayesian net • What is AgenaRisk

3. Software Measurement: Uncertainty and Causal Modeling P 2/12 Causes leading to project failure • Starting with inadequate budget or schedule • Starting without adequate committed resources • Starting without occurring of requirements flowdown • Major project designs before understanding the relation between cost, performance, schedule and risk

4. Software Measurement: Uncertainty and Causal Modeling P 3/12 Software Measurement and Analysis • How well are we meeting schedules and budgets? • Has our performance really improved? • What software practices and/or technologies should our organization invest in? • How does our organization's performance compare to other organizations' performances?

5. Software Measurement: Uncertainty and Causal Modeling P 4/12 Benefits of Software Measurement • Encourages appropriate behavior • Pinpoints opportunities for improvement • Leverage resources • Enhances requirements process • Manage workloads • Only measuring can reduce cost by 15-20 %

6. Software Measurement: Uncertainty and Causal Modeling P 5/12 Uncertainty and Software Measurement • Roles of software measurement • Quality control • Effort estimation • Uncertainty • Definition • How does it relate to software management

7. Software Measurement: Uncertainty and Causal Modeling P 6/12 Risk Management • Risk: Probability of not achieving a defined project goal • Risk Management: Act of dealing with risk • Assessing risk issues • Developing risk handling strategies • Monitoring risk

8. Software Measurement: Uncertainty and Causal Modeling P 7/12 Software Quality • External product view (quality-in-use) • Defined by measuring external properties • Internal product view • Quality of intermediate products

9. Software Measurement: Uncertainty and Causal Modeling P 8/12 Prerequisites to produce a defect free product • Good development process • Well qualified staff working on a defined specification • … Are the above enough? To ensure that the process conditions are right, for example checking intermediate products

10. Software Measurement: Uncertainty and Causal Modeling P 9/12 Causal Relationships in software Development • Simpson’s paradox

11. Software Measurement: Uncertainty and Causal Modeling P 10/12 Causal Relationships in Software Development In software engineering causal relationships are rarely straight forward 60 10 30 40 0

12. p(D = high | T = high) p(T = high) 0.8 X 0.2 = = 0.32 p(T = high | D = high) = 0.5 p(D = high) Software Measurement: Uncertainty and Causal Modeling P 11/12 Causal Modeling and Bayesian Method

13. Software Measurement: Uncertainty and Causal Modeling P 12/12 Summary • Quality measurement is an inevitable in software management • Risk management covers software quality • Modeling risk helps us to have a better understanding of project risks • Combining causal modeling and Bayesian method gives an effective risk management approach • An introduction to AgenaRisk tool

14. Thanks and Questions