Singular

Hello,

what is a good method to detect the error rate (defects per 1.000 loc) of a program?

I'm aware of two methods:


1. needs data from two testers A and B which independently analysed the same code parts. If tester A observed m errors and tester B n errors, whereas
k of the errors were observed by both testers, a rough estimate is (m*n) / k

problem: get data from at least two observers

2. artificially insert m bugs into source code analysed by testers. Assume x bugs were detected in total and z of them were artificial bugs,
then the estimate is (x-z)*(m/z)


Jakob