p5 prereg (4 Nov 05:35, before looking): Uncorrected mercury barometers read ~0.0027 inHg high per degF warmer. Same-place pairs share temperature, so a ship only shows a temperature signal relative to partners if its correction practice differs from theirs.
Predictions: (a) >=5 of ~100 ships show significant (p<0.01) dependence of their partner-based estimate on |lat| (warm-proxy); (b) among the 10 stepping drifters, the step mostly disappears (|step| halves for >=6) once |lat| is regressed out; (c) sign of slopes mixed across ships (some corrected, some not), not all one way.
