For my PhD dissertation, I developed a method for identifying significant genetic markers for cervical cancer. It is part of a larger project on the subject. This ensemble of projects makes an actual difference to actual women who can receive intervention care much sooner than they would have previously.

Could some one with only a bachelor's degree have developed these methods? Yes. As long as they understood Bayesian analysis at a doctoral level. And they would likely need some multivariate real analysis under their belt, such as would be taught in Maths 541 and 641 at BYU. You might need to have a good grasp of manifold and topology, so throw some other graduate math classes onto the pile. Since this bachelor's degree person has foregone a graduate degree, they are now self teaching all of these subjects. Let's hope they learned them correctly! And let's hope that their employer is not paying them to learn these things, because that would not be economical at all. (Would you rather pay me $20k/year stipend to earn a PhD, or $60k/year salary to self learn graduate level statistics and math with only a BSc.?)

By the time someone has self learned (the esteemed on the job training) everything needed to develop the methods my team created, they might as well have just gotten the PhD. Earning a PhD is much more than just writing a big paper and talking about it. It is the process. And that process is much more economically achieved by getting a PhD than it is by having employees self learn on the job.