STATSTISTICS

 

Last week, you examined how operating system, database management system, and programming language affected work load per function point. This would provide useful information for project management, but there is not yet enough information to get a work load estimated based on function point analysis. This week, you will utilize regression analysis to develop this estimate.

First, you will investigate the overall relationship between work hours and point analysis. Start in Excel which is great for graphing and basic analysis.

1. Create a scatter plot in Excel plotting work hours on the y-axis and function point count on the x-axis. Copy this chart to your overall document. Explain what trends you notice looking at the chart.
2. Now, you can investigate the trend with regression and correlation. Right click on one of the data points on your graph and select Add Trendline.
   1. Consider which trends might be logical for this situation. A linear trend would mean every function point required an average number of work hours to complete. Exponential would mean that each function point in a large project would take more time than in a small project. Logarithmic would mean that each function in a large project would take less time than in a small project. Polynomial or Power would indicate a complex dependence. Which of these do you think might model this situation? Why?
   2. Select the options to Display Equation on chart and Display R-squared value on chart. The R-square value describes how well the trend line fits the data. Click through the different regression types and find which has the best (highest) R-squared value. Identify the regression type and record the equation and the R-squared value in your overall document.
   3. If you made your polynomial higher order, the fit would get even better, but that does not necessarily make it a good model. The complex relationship does not logically fit the situation. A linear relationship is more logically explained. Find the best fit linear equation and record it along with the R-square value. What do the slope and intercept each tell you about work hours and function points.
   4. Presumably, a project with no function points would not require any work hours. You can model that situation by going back into your linear trend line settings and selecting Set Intercept = 0. Record the resulting equation and R-squared value. What does this tell you about the relationship between work hours and function points?
3. Although you now have a model for determining work hours based on the number of function points, it is not the best model possible. You know from your previous analysis that there are significant differences for different operating system, database management system, and language. Taking these into account will support a better model. Describe how you might take these nominal variables into account in creating a better model.
4. Estimate how many work hours would be required at this facility to complete a 1000 function point project on Unix, IDMS, and Cobol. Explain how you arrived at this estimate. Describe how that estimate could be useful to the facility.
5. For a project done at another facility by another group, would your estimate based on results at this facility hold true? Why or why not? What factors might influence the comparability of two software teams?