Chapter 8.4, Problem 8.58ACI

Predicting software defects. Refer to the PROMISE Software Engineering Repository data on 498 modules of software code written in “C” language for a NASA spacecraft instrument, saved in the file. (See Exercise 3.132, p. 185). Recall that the software code in each module was evaluated for defects; 49 were classified as “true” (i.e., module has defective code), and 449 were classified as “false” (i.e., module has correct code). Consider these to be independent random samples of software code modules. Researchers predicted the defect status of each module using the simple algorithm, “If number of lines of code in the module exceeds 50, predict the module to have a defect.” The accompanying SPSS printout shows the number of modules in each of the two samples that were predicted to have defects (PRED_LOC = _“yes”) and predicted to have no defects (PRED_LOC = _“no”). Now, define the accuracy rate of the algorithm as the proportion of modules that were correctly predicted. Compare the accuracy rate of the algorithm when applied to modules with defective code with the accuracy rate of the algorithm when applied to modules with correct code. Use a 99% confidence interval.

DEFECT PRED_LOC Crosstabulation

Count

		PRED_LOC		Total
		no	yes
DEFECT	false	400	49	449
	true	29	20	49
total		429	69	496