Test whether three-hour bins can identify a solar-wind lag analysishelp-wantednegative-control

Depends on task 1 producing documented overlapping measurements and resolved interval labels. Test how temporal resolution and missingness change an apparent solar-wind/Kp lag; this is an identifiability exercise, not a causal forecast.

Before seeing correlations freeze one summary per three-hour Kp bin, for example the median southward GSM Bz magnitude with valid sample count, and a lag grid {0, 3, 6 hours}. State the sign convention, bin boundaries and spacecraft propagation assumption. Compare this to the deliberately naive result that forward-fills each Kp value into minute samples; label the latter a negative-method control and never count those repeated values as independent observations.

Acceptance: show raw bin count, coverage thresholds and excluded bins; publish the aligned table and analysis recipe. Use whole contiguous time blocks for any held-out evaluation; a seven-day exploratory window cannot establish a general physical lead time. Report all three lags without picking a hidden best value. Check a block time-shift null and sensitivity to omitted gaps and adjacent bin conventions; make no calibrated significance claim without enough independent blocks. If interval uncertainty, too few bins or changing spacecraft prevents interpretation, say exactly why the lag is unidentifiable.

Budget: reuse the frozen dataset under 10 MB, CPU only, no ML/GPU; record peak memory and runtime. Do not claim local aurora visibility from planetary Kp. Reviewer should be able to overturn the result by reproducing the table with a defensible timestamp convention.