7,613
edits
Changes
Jump to navigation
Jump to search
no edit summary
**Agglomeration!
**Note: Needs updating to reflect latest decisions
===BuildBaseTables===
To this end, the base tables, which are used by multiple different projects, are built in BuildBaseTables.sql
Build notes:
*Common code from MatchingEntrepsV3.sql was moved to BuildBaseTables.sql
*Almost all of the women's code now appears to be in the new base
*The super tables still need building
*The code for the Matching VC-Entreps project is now almost entirely in RevisedDBaseCode.sql
==Agglomeration==
A new Ranking.sql file was created in E:\projects\vcdb4 and run.
There are imperfect matches between cities in VentureXpert and places in Tiger. In the past, this hasn't mattered as we'd considered such a small sub set of cities (and had gotten lucky on coverage). Geocoding covers 97.2607% (47.760 out of 49,136) of companybasecore, 48,815 (99.35%) of which have a non-null city. This drops to 96.42% of firms with datefirstinv >= 2010. Using just geocoded would therefore cost us a small number of firms, but these may not be randomly distributed. What we need, then is a complete lookup table...
