This is in reference to the "Stopping Power" books that were the rage in the 90s.
1. The database was WAY too small. The conclusions drawn from such a small database couldn't be viewed as terribly meaningful.
2. The data favored cartridges used in more accurate platforms. In other words, .45 ACP was typically fired from full-sized duty weapons, capable of more precise fire. However, guns like the .25 and .32 are usually fired from back-up, concealed guns. This uninentionally mixed variables, which is the biggest no-no in a "scientific" test.
3. The whole notion of a one-shot stop is flawed. What caused the "stop?" How immediate was it? Was it a one-shot stop only because the shooter wasn't accurate enough to get in a second hit or because his/her gun jammed? How big was the psychological effect (which in humans can range from none to fainting just at the sight of muzzle flash)? Did a shot in the upper thigh count as a one-shot? It's just a bad idea that's difficult to correlate to a cartridge's "stopping power."
4. The numbers were too good to be true. Statisticians who viewed the data often concluded that it looked faked...pre-conceived notions led to a skewing of numbers that promoted certain cartridge/bullet/velocity combinations.
You'll notice that the most accepted terminal performance testing involved today looks at observed data...depth of penetration, diameter of expanded bullet, % fragmentation, etc. However, you'll NEVER see anybody worth two licks try to correlate that observed data to a magical number such as "one-stop shot percentage." The best you'll get is a "this is how this particular round performs through the various FBI criteria tests." It's your job to conclude what you will from that.