The not-shocking conclusion of team of statisticians is if we reduce our sample size/reduce the data collection on the nuclear stockpile, we’ll have reduced confidence in its ability to operate as designed.
"It’s commonly suggested that we turn off data collection or slow down data collection. When you are fairly certain that you have an aging stockpile, our study shows that ignoring the problem by not collecting data is a huge mistake," [Brigham Young University statistics professor Shane] Reese said.
That statement makes me feel better about my understanding of statistical theory.
Reese and his team determined that should the federal government reduce arsenal monitoring by 50 percent, the reliability that the weapon would operate as intended would fall by 15 percent to 80 percent over a seven-year period. Should monitoring be halted entirely, confidence in a weapon’s performance would fall to close to 50 percent.
Better stated is this version:
If the government cut monitoring in half, confidence in the system’s reliability would drop from 95 to 80 percent within seven years. And if monitoring stopped altogether, the reliability would drop to almost the same odds that you’d find with flipping a coin.