You can use this Splitfile to rebuild the image. Read about inspecting provenance in the documentation.
SQL {CREATE TABLE qoz_vote_fraction
AS WITH county_tracts AS (SELECT count(*) AS total_tract_count
, pg_catalog.substring(lpad("TractId"::text
, 11
, '0')
, 0
, 6) AS county_id
FROM "splitgraph/census:17a596cf4642f50713a817dcd34677cec329e3a83b884a38e115389ecc7e1f0e".acs2017_census_tract_data
GROUP BY county_id)
, qoz_tracts AS (SELECT count(*) AS qoz_tract_count
, pg_catalog.substring(lpad("Census Tract Number"::text
, 11
, '0')
, 0
, 6) AS county_id
FROM "splitgraph/qoz:5818ea35bf99c1f93fb3fb4f751c69b93db1e6a1354b2fc558f8a0036041588a".qoz
GROUP BY county_id)
, by_candidate AS (SELECT lpad(county_fips::text
, 5
, '0') AS county_id
, sum(CASE
WHEN (candidate_normalized = 'trump')
THEN votes
ELSE 0
END) AS trump_votes
, sum(CASE
WHEN (candidate_normalized = 'clinton')
THEN votes
ELSE 0
END) AS clinton_votes
, sum(votes) AS total_votes
FROM "splitgraph/2016_election:3835145ada3f07cad99087d1b1071122d58c48783cbfe4694c101d35651fba90".precinct_results
GROUP BY county_id)
SELECT v.county_id
, COALESCE((q.qoz_tract_count)::numeric / (t.total_tract_count)::numeric
, 0.0) AS qoz_tract_fraction
, trump_votes / total_votes AS trump_vote_fraction
, clinton_votes / total_votes AS clinton_vote_fraction
, total_votes
FROM by_candidate AS v
LEFT JOIN county_tracts AS t ON t.county_id = v.county_id
LEFT JOIN qoz_tracts AS q ON t.county_id = q.county_id}