Question: Import the data set “project.xls”. (Launch SPSS and “Open an existing data source”. Change the desired file type from SPSS (*.sav) to Excel. Open “project.xls”. Have it “Read variable names”. Save the data file as “project.sav”.)
This file contains summary project data, including the size of the project (in function points), the overall duration or schedule of the project (in calendar months), and the total amount of effort needed to develop the project (in staff months).

a) Plot Schedule (Y) versus Effort (X) using a simple scatter plot.

b) Generate a linear and power regression analysis for Schedule (dependent) vs. Effort (indep.). (Use Analyze/Regression/Curve Estimation. Choose Linear and Power Models, and select “Display ANOVA table”.) The power regression finds the best fit for equations of the form
(dependent variable Y) = (Const.)*(indep. variable X)^(B)
where “B” is the value of B for the variable X, and “Const.” is the value of B for the row labeled “Const.”

c) Create a table showing the R Square and Standard Error for both Linear and Power models.

d) Choose the best-fitting model for Effort versus Schedule. Write the equation which best models the relationship between effort and schedule, including standard error terms.

e) Prove that the 95% confidence interval for the two constants in the best-fitting model does not include zero. Why is this an important thing to check?

f) Based on the curve fit graph, would you trust the model chosen in part d) for small projects? For large projects? Explain your answers.

^g) In McConnell, Rapid Development, a relationship between schedule estimate in months and total staff months expended on a software project is given by:
Schedule = 3.0 x (Effort)^1/3
How does your conclusion from part d) compare to this result?
More specifically, do the values of your coefficients agree with McConnell’s, to a 95% level of confidence? Show your work.