search

COVID19Tracking / issues

Repo for issues and administration
45 stars 6 forks source link

[NM] Shift timeseries by a day and patch a few values that do not match state reporting date #961

Closed space-buzzer closed 3 years ago

space-buzzer commented 3 years ago

State or US: New Mexico Problem NM updates their dashboard late during the day, so up until now, we were capturing values reported on the previous day. Starting today (Nov-18, 2020) we will be capturing the values NM publishing on the day of their publish, such that our dating of the data matches the states dating. For this, we will shift the entire time series by 1 day.

NM publishes historic time series, available at: https://cvprovider.nmhealth.org/public-dashboard.html (clicking "Show Historical Statewide Data). We'll use this to also patch a few days and fill missing values at the beginning of the series.

BrettBoval commented 3 years ago

For what it's worth, NM also changed their methodology for collecting hospitalization data which resulted in a large increase (~40%) in reported values this week. (https://www.santafenewmexican.com/news/coronavirus/new-mexico-sets-daily-record-for-covid-19-cases-deaths/article_7c7cfb58-2929-11eb-b562-a39d722bccdf.html).

image

I don't have an obvious solution on how to rectify this disconnect in the timeseries, but at the very least I think it's worth adding a note at https://covidtracking.com/data/state/new-mexico.

space-buzzer commented 3 years ago

Yeah, this is a going to be a separate issue (at least a note). Since they didnt backfill their historic timeseries of hospitalizations (+ it matches what we've been capturing in the past), I'll use their history for the small fixes for days where we've been capturing the "correct" day.

space-buzzer commented 3 years ago

Patching the time series for the few days that were now reporting the future.

Multi-edit tool-PRODUCTION - NM-reb-20201118.csv.txt Multi-edit tool-PRODUCTION - NM-reb-20201118_post.csv.txt