https://phys.org/news/2018-07-billions-dollars-aids.html

Unfortunately, PEPFAR didn't create a statistical plan at the outset that could easily answer these questions. A statistical plan is a bit like asking your accountant what data you should record before you start your business—so you'll know whether you're making or losing money. The question facing PEPFAR was whether they had collected enough data over the years to evaluate their program interventions.

These kinds of after-the-fact analyses are fraught with difficulty. "Planned evaluations are preferable, of course, but we can still learn a lot from post-hoc evaluations," says Donna Spiegelman, the newly appointed Susan Dwight Bliss Professor of Biostatistics at the Yale School of Public Health. "I would never let the perfect be the enemy of the good."