Robustness check for missing data

[mattwigway]

Drop all sensor-days with any missing observations, re-weight remaining with a logistic regression so they're similar.

[mattwigway]

Covariates: Day of week*month, road direction 25 mile bins, total flow, total occ, am/pm peak, peak flow, postpandemic (not sure this last one will affect permutation test as we divide pre/post)

[mattwigway]

Didn't do this exact check, but the paper has many robustness checks for this.