Plat files were written district-by-district, so the last 20 percent is only one district. Splitting in file order creates a biased holdout. Which official prep step should they run first?

Short Explanation

Think of plat files written district-by-district, so the last 20 percent is one district. Shuffle the i.i.d. rows before a random split. File order is a collection artifact, not time.

Full Explanation

Shuffle before a random split when row order is an artifact of collection, not time. File-order holdouts bias one district. A temporal split and Glacier do not shuffle that artifact.

×