earth-the-technique-grid-needs-values-in-its-cellslisted

# Earth: a cell count is not a cell value You will draw the grid. An archive stratified by measurement technique invites a technique-by-unit matrix almost at once, and it is the right picture — but the first version of it holds how many files fall in each cell, and that version answers a filing question. The deliverable is the same grid with the *estimate* in each cell. Draw both, in that order, and never let the first stand in for the second. The two look almost identical — same rows, same columns, same annotated cells, same marginals — which is exactly why the substitution goes unnoticed. Nothing internal flags it, because the picture that was planned did get drawn, on time, under its planned filename. What the value grid has to *say* once it exists is not this skill's subject and is already written down: `earth-inter-technique-spread-is-a-result-not-a-caveat` for the reference technique, each technique's own observing window, rate against interannual variability, what to do when a statistic sits on a different population from your arms, and where the offsets have to be printed; `earth-report-the-lattice-and-show-the-field` for the lattice it is drawn over. Follow both. This skill is the two failures *upstream* of them — the two ways a run arrives at analysis with no per-technique values to put anywhere. ## Read every column header before you write any pipeline code Open the header line of every supplied file, including the ones in a directory whose name says results r