Changes

{{Project|Has project output=Data,How-to|Has sponsor=McNair ProjectsCenter

The successors to this project include:*[[Augi Liebster (Work Log)VCDB24]], which is the most recent iteration.*[[VCDB23]]*[[VCDB20Q3]]*[[VCDB20H1]]*[[VCDB4]]

E:\McNair\Projects\VentureXpertDatabase\ScriptsForSDCExtract

E:\McNair\Projects\VentureXpertDatabase\ExtractedDataQ2

My script scripts for loading tables and data is are in one big text file in the location: E:\McNair\Projects\VentureXpertDatabase\vcdb3\LoadingScripts

disclosedamt moneyreal, investor fundname varchar(255)

LEFT JOIN MANoDups AS C ON A.targetnametargetnamestd=C.targetname AND A.targetstate=C.targetstate AND A.announceddate=C.announceddate; --7171

SELECT A.coname, A.statecode, A.datefirstinv, A.datelastinv, A.city, B.ipovsma, CASE WHEN aB.mavsipo=0 THEN 0::int WHEN a.mavsipo=1 ipovsma IS NOT NULL THEN 1::int ELSE NULL0::int END AS mavsipoExit, CASE WHEN aB.mavsipoipovsma=0 1 THEN BC.announceddate proceedsamt::numeric WHEN a.mavsipoipovsma=1 0 THEN CD.issuedate transactionamt::numeric ELSE NULL::date numeric END AS exitdateexitvaluem, CASE WHEN aB.mavsipoipovsma=0 OR a1 THEN C.mavsipoissuedate WHEN ipovsma=1 0 THEN 1::int D.announceddate ELSE 0NULL::int date END AS exitedexitdate, CASE WHEN aB.mavsipoipovsma=0 1 THEN Bextract(year from C.transactionamt issuedate) WHEN a.mavsipoipovsma=1 0 THEN Cextract(year from D.proceedsamt announceddate) ELSE NULL::money int END AS exitvalueexityear FROM ExitKeysClean companybasecore AS A LEFT JOIN MANoDups ExitKeys AS B ON A.targetnameconame=B.targetname coname AND A.targetstatestatecode=B.targetstate statecode AND A.announceddatedatefirstinv=B.announceddatedatefirstinv LEFT JOIN IPONoDups IPOCleanNoDups AS C ON A.issuernameconame=C.issuer coname AND A.issuerstatestatecode=C.statecode AND A.issuedatedatefirstinv=C.issuedatedatefirstinv LEFT JOIN MACleanNoDups AS D ON A.coname=D.coname AND A.statecode=D.statecode AND A.datefirstinv=D.datefirstinv;

A helpful page here is the [[Geocode.py]] page which explains how to use the Geocoding script. You will have to tweak the Geocode script when geocoding as each of these tables has a different primary key. It is vital that you include the primary keys in the file you input and output from the Geocoding script. Without these, you will not be able to join the latitudes and longitudes back to the firm, branch office, or company base tables. Geocoding costs money since we are using the Google Maps API. The process doesn't cost much, but in order to save money I tried to salvage as much of the preexisting geocode information I could find.===Companies===I found the table of old companies with latitudes and longitudes in vcdb2 and loaded these into vcdb3.  DROP TABLE oldgeocords; CREATE TABLE oldgeocords ( coname varchar(255), statecode varchar(10), datefirstinv date, ivestedk real, city varchar(255), addr1 varchar(255), addr2 varchar(100), latitude numeric, longitude numeric );  \COPY oldgeocords FROM 'companybasegeomaster.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --44740 The API occasionally will give erroneous latitude and longitude readings. In order to catch only the good ones, I found the latitude and longitude lines that encompass the mainland US and created an exclude flag to make sure companies were in this box. I then created flags to include companies in Puerto Rico, Hawaii, and Alaska. Companies that were in these places often had wrong latitude and longitude readings of 44.93, 7.54, so I ran a query making sure that these weren't listed.   DROP TABLE geoallcoords; CREATE TABLE geoallcoords AS SELECT *, CASE WHEN longitude < -125 OR longitude > -66 OR latitude < 24 OR latitude > 50 OR latitude IS NULL OR longitude IS NULL THEN 1::int ELSE 0::int END AS excludeflag FROM oldgeocords; --44740  DROP TABLE geoallcoords1; CREATE TABLE geoallcoords1 AS SELECT *, CASE WHEN statecode='PR' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as prflag, CASE WHEN statecode='HI' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as hiflag, CASE WHEN statecode='AK' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as akflag FROM geoallcoords; --44740 I then included only companies that were either in the mainland US, Hawaii, Alaska, or Puerto Rico.   DROP TABLE goodgeoold; CREATE TABLE goodgeoold AS SELECT A.*, B.latitude, B.longitude, B.prflag, B.excludeflag, B.hiflag, B.akflag FROM companybasecore AS A LEFT JOIN geoallcoords1 AS B ON A.coname=B.coname AND A.statecode=B.statecode AND A.datefirstinv=B.datefirstinv WHERE excludeflag=0 or prflag=1 or hiflag=1 or akflag=1; --38498 I then found the remaining companies that needed to be geocoded. Only companies that have addresses listed are able to be accurately geocoded. If we attempt to geocode based on city, the location returned will simply be the center of the city. Thus, I chose the companies that we did not already have listings for and had a valid address.  DROP TABLE remaininggeo; CREATE TABLE remaininggeo AS SELECT A.coname, A.statecode, A.datefirstinv, A.addr1, A.addr2, A.city, A.zip FROM companybasecore AS A LEFT JOIN goodgeoold AS B ON A.coname=B.coname AND A.statecode=B.statecode AND A.datefirstinv=B.datefirstinv WHERE B.coname IS NULL AND A.addr1 IS NOT NULL; --5955  \COPY remaininggeo TO 'RemainingGeo.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --5955 I copied this table into excel to concatenate the address, city, state, and zipcode columns into one column. This can and should be done in SQL, but I was not aware this could be done. I then ran remaininggeo through the Geocode script with columns coname, statecode, datefirstinv, and address in the inputted file.  DROP TABLE remaining; CREATE TABLE remaining ( coname varchar(255), statecode varchar(10), datefirstinv date, latitude numeric, longitude numeric );  \COPY remaining FROM 'RemainingLatLong.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --5955 I then ran the same geographical checks on the newly geocoded companies and found all of the good geocodes.   DROP TABLE geoallcoords2; CREATE TABLE geoallcoords2 AS SELECT *, CASE WHEN longitude < -125 OR longitude > -66 OR latitude < 24 OR latitude > 50 OR latitude IS NULL OR longitude IS NULL THEN 1::int ELSE 0::int END AS excludeflag FROM remaining; --5955  DROP TABLE geoallcoords3; CREATE TABLE geoallcoords3 AS SELECT *, CASE WHEN statecode='PR' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int 0::int END as prflag, CASE WHEN statecode='HI' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as hiflag, CASE WHEN statecode='AK' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as akflag FROM geoallcoords2; --5955  DROP TABLE goodgeonew; CREATE TABLE goodgeonew AS SELECT A.*, B.latitude, B.longitude, B.prflag, B.excludeflag, B.hiflag, B.akflag FROM companybasecore AS A LEFT JOIN geoallcoords3 AS B ON A.coname=B.coname AND A.statecode=B.statecode AND A.datefirstinv=B.datefirstinv WHERE excludeflag=0 or prflag=1 or hiflag=1 or akflag=1; --5913 I then combined the old and new geocodes and matched them back to the company base table to get a geo table for companies.  DROP TABLE geocodesportco; CREATE TABLE geocodesportco AS SELECT A.* from goodgeonew UNION SELECT B.* from goodgeoold; --44411  DROP TABLE portcogeo; CREATE TABLE portcogeo AS SELECT A.*, B.latitude, B.longitude FROM companybasecore AS A LEFT JOIN Geocodesportco AS B ON A.coname=B.coname AND A.datefirstinv=B.datefirstinv AND A.statecode=B.statecode; --48001 ===Firms===This process is largely the same as for companies. I found old firms that had already been geocoded and checked for accuracy.  DROP TABLE oldfirmcoords; CREATE TABLE oldfirmcoords ( firmname varchar(255), latitude numeric, longitude numeric ); \COPY oldfirmcoords FROM 'FirmCoords.txt' DELIMITER AS E'\t' HEADER NULL AS '' CSV --5556  DROP TABLE firmoldfilter; CREATE TABLE firmoldfilter AS SELECT *, CASE WHEN longitude < -125 OR longitude > -66 OR latitude < 24 OR latitude > 50 OR latitude IS NULL OR longitude IS NULL THEN 1::int ELSE 0::int END AS excludeflag FROM oldfirmcoords; --5556 Since oldfirmcoords does not have state codes, we have to find a way to include state codes to add in companies based in Puerto Rico, Hawaii, and Alaska. I did this by matching the firmoldfilter table back to the firm base table. I used the coalesce function because we wanted to exclude companies that we had not geocoded due to faulty addresses.   DROP TABLE firmcoordsmatch1; CREATE TABLE firmcoordsmatch1 AS SELECT A.firmname, A.state, B.latitude, B.longitude, COALESCE(B.excludeflag, 1) AS excludeflag FROM firmbasecore AS A LEFT JOIN firmoldfilter AS B ON A.firmname=B.firmname; --15437 Then the process of tagging the PR, HI, and AK companies and including only correctly tagged companies is the same as for companies.   DROP TABLE firmcoordsexternal; CREATE TABLE firmcoordsexternal AS SELECT *, CASE WHEN state='PR' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as prflag, CASE WHEN state='HI' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as hiflag, CASE WHEN state='AK' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as akflag FROM firmcoordsmatch1; --15437  DROP TABLE goodfirmgeoold; CREATE TABLE goodfirmgeoold AS SELECT A.*, B.latitude, B.longitude, B.prflag, B.excludeflag, B.hiflag, B.akflag FROM firmcoreonedupremoved AS A LEFT JOIN firmcoordsexternal AS B ON A.firmname=B.firmname WHERE excludeflag=0 or prflag=1 or hiflag=1 or akflag=1; --5346 Find the remaining firms and run the geocode script on these firms  DROP TABLE remainingfirm; CREATE TABLE remainingfirm AS SELECT A.firmname, A.addr1, A.addr2, A.city, A.state, A.zip FROM firmcoreonedupremoved AS A LEFT JOIN goodfirmgeoold AS B ON A.firmname=B.firmname WHERE B.firmname IS NULL AND A.addr1 IS NOT NULL AND A.msacode!='9999'; --706  \COPY remainingfirm TO 'FirmGeoRemaining.txt' DELIMITER AS E'\t' HEADER NULL AS '' CSV --706  DROP TABLE firmremainingcoords; CREATE TABLE firmremainingcoords( firmname varchar(255), latitude numeric, longitude numeric );  \COPY firmremainingcoords FROM 'FirmRemainingCoords.txt' DELIMITER AS E'\t' HEADER NULL AS '' CSV --706 Follow the same filtering process as above to get the good geocodes.   DROP TABLE firmnewfilter; CREATE TABLE firmnewfilter AS SELECT *, CASE WHEN longitude < -125 OR longitude > -66 OR latitude < 24 OR latitude > 50 OR latitude IS NULL OR longitude IS NULL THEN 1::int ELSE 0::int END AS excludeflag FROM firmremainingcoords; --706  DROP TABLE firmcoordsmatch2; CREATE TABLE firmcoordsmatch2 AS SELECT A.firmname, A.state, B.latitude, B.longitude, COALESCE(B.excludeflag, 1) AS excludeflag FROM firmcoreonedupremoved AS A LEFT JOIN firmnewfilter AS B ON A.firmname=B.firmname; --15437  DROP TABLE firmcoordsexternalremaining; CREATE TABLE firmcoordsexternalremaining AS SELECT *, CASE WHEN state='PR' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as prflag, CASE WHEN state='HI' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as hiflag, CASE WHEN state='AK' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as akflag FROM firmcoordsmatch2; --15437  DROP TABLE goodfirmgeonew; CREATE TABLE goodfirmgeonew AS SELECT A.*, B.latitude, B.longitude, B.prflag, B.excludeflag, B.hiflag, B.akflag FROM firmcoreonedupremoved AS A LEFT JOIN firmcoordsexternalremaining AS B ON A.firmname=B.firmname WHERE excludeflag=0 or prflag=1 or hiflag=1 or akflag=1; --703 Combine the old and new geocoded firms and match them to firm base to get a firm geo table.  DROP TABLE firmgeocoords; CREATE TABLE firmgeocoords AS SELECT * FROM goodfirmgeonew UNION SELECT * FROM goodfirmgeoold; --6049  DROP TABLE firmgeocore; CREATE TABLE firmgeocore AS SELECT A.*, B.latitude, B.longitude FROM firmbasecore AS A LEFT JOIN firmgeocoords AS B ON A.firmname=B.firmname; --15437 ===Branch Offices===I did not use old branch office data because I could not find it anywhere in the old data set. I have since found old data in the table firmbasecoords in vcdb2.  First copy all of the needed data out of the database to do geocoding.  \COPY (SELECT A.firmname, A.boaddr1, A.boaddr2, A.bocity, A.bostate, A.bozip FROM bonound AS A WHERE A.boaddr1 IS NOT NULL) TO 'BranchOffices.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --2046 Then load the data into the database and follow the same filtering process as above.  DROP TABLE bogeo; CREATE TABLE bogeo ( firmname varchar(255), latitude numeric, longitude numeric );  \COPY bogeo FROM 'BranchOfficesGeo.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --2046  DROP TABLE bogeo1; CREATE TABLE bogeo1 AS SELECT *, CASE WHEN longitude < -125 OR longitude > -66 OR latitude < 24 OR latitude > 50 OR latitude IS NULL OR longitude IS NULL THEN 1::int ELSE 0::int END AS excludeflag FROM bogeo; --2046  DROP TABLE bomatchgeo; CREATE TABLE bomatchgeo AS SELECT A.*, B.latitude, B.longitude, COALESCE(B.excludeflag, 1) AS excludeflag FROM branchofficecore AS A LEFT JOIN bogeo1 AS B ON A.firmname=B.firmname; --10032  DROP TABLE bogeo2; CREATE TABLE bogeo2 AS SELECT *, CASE WHEN bostate='PR' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as prflag, CASE WHEN bostate='HI' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as hiflag, CASE WHEN bostate='AK' AND latitude!=44.9331 AND longitude!=7.54012 THEN 1::int ELSE 0::int END as akflag FROM bomatchgeo; --10032 Match the correctly geocoded branch offices back to firm base to get the final table.  DROP TABLE bogeocore1; CREATE TABLE bogeocore1 AS SELECT * FROM bogeo2 WHERE excludeflag=0 or prflag=1 or hiflag=1 or akflag=1; --1161  DROP TABLE firmbogeo; CREATE TABLE firmbogeo AS SELECT A.*, B.latitude AS BOLatitude, B.longitude AS BOLongitude FROM firmgeocore AS A LEFT JOIN bogeocore1 AS B ON A.firmname=B.firmname; --15437 ==Creating People Tables==We pulled data on executives in both portcos and funds. I describe the process below. If any of the explanations don't make sense, I also describe most tables in the section called Marcos's Code.===Company People=== DROP TABLE titlelookup; CREATE TABLE titlelookup( fulltitle varchar(150), charman int, ceo int, cfo int, coo int, cio int, cto int, otherclvl int, boardmember int, president int, vp int, founder int, director int );  \COPY titlelookup FROM 'Important Titles in Women2017 dataset.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --628 This table lists various titles one can have and identifies where they fall under traditional executive titles.  DROP TABLE copeople; CREATE TABLE copeople( datefirstinv date, cname varchar(150), statecode varchar(2), prefix varchar(5), firstname varchar(50), lastname varchar(50), jobtitle varchar(150), nonmanaging varchar(1), prevpos varchar(255) );  \COPY copeople FROM 'Executives-NoFoot-normal.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --194359 This table gets various executives from portcos. This is loaded from SDC. Next we have to identify which traditional executive level job the listed job title corresponds to. It also identifies whether a prefix identifies an executive as male or female. I made a stupid mistake of writing cname instead of coname when loading in the data. If you want to save yourself work, write coname.  DROP TABLE copeoplebase; CREATE TABLE copeoplebase AS SELECT copeople.*, CASE WHEN prefix='Ms' THEN 1::int WHEN prefix='Mr' THEN 0::int ELSE Null::int END AS titlefemale, CASE WHEN prefix='Ms' THEN 0::int WHEN prefix='Mr' THEN 1::int ELSE Null::int END AS titlemale, CASE WHEN prefix='Dr' THEN 1::int ELSE 0::int END AS doctor, CASE WHEN prefix IS NULL THEN 0::int ELSE 1::int END AS hastitle, CASE WHEN prefix IS NULL AND firstname IS NULL AND lastname IS NULL THEN 0::int ELSE 1::int END AS hasperson, CASE WHEN fulltitle IS NOT NULL THEN 1::int ELSE 0::int END AS hastitlelookup, CASE WHEN charman IS NOT NULL THEN charman ELSE 0::int END AS chairman, CASE WHEN ceo IS NOT NULL THEN ceo ELSE 0::int END AS ceo, CASE WHEN cfo IS NOT NULL THEN cfo ELSE 0::int END AS cfo, CASE WHEN coo IS NOT NULL THEN coo ELSE 0::int END AS coo, CASE WHEN cio IS NOT NULL THEN cio ELSE 0::int END AS cio, CASE WHEN cto IS NOT NULL THEN cto ELSE 0::int END AS cto, CASE WHEN otherclvl IS NOT NULL THEN otherclvl ELSE 0::int END AS otherclvl, CASE WHEN boardmember IS NOT NULL THEN boardmember ELSE 0::int END AS boardmember, CASE WHEN president IS NOT NULL THEN president ELSE 0::int END AS president, CASE WHEN vp IS NOT NULL THEN vp ELSE 0::int END AS vp, CASE WHEN founder IS NOT NULL THEN founder ELSE 0::int END AS founder, CASE WHEN director IS NOT NULL THEN director ELSE 0::int END AS director FROM copeople LEFT JOIN titlelookup ON copeople.jobtitle=titlelookup.fulltitle; --194359 Next we will try to identify whether an executive is male or female based on their names.  DROP TABLE namegender; CREATE TABLE namegender AS SELECT firstname, CASE WHEN countfemale > 0 AND countmale=0 THEN 1::int ELSE 0::int END AS exclusivelyfemale, CASE WHEN countmale > 0 AND countfemale=0 THEN 1::int ELSE 0::int END AS exclusivelymale FROM (SELECT firstname, COALESCE(sum(titlefemale),0) as countfemale, COALESCE(sum(titlemale),0) as countmale FROM copeoplebase WHERE doctor=0 GROUP BY firstname) As T WHERE NOT (countfemale > 0 AND countmale>0); --12736 The next table expands CoPeopleBase to include information on executive gender and executive position.  DROP TABLE CoPeopleFull; CREATE TABLE CoPeopleFull AS SELECT copeoplebase.*, CASE WHEN titlefemale=1 THEN 1::int WHEN exclusivelyfemale=1 THEN 1::int ELSE 0::int END AS female, CASE WHEN titlemale=1 THEN 1::int WHEN exclusivelymale=1 THEN 1::int ELSE 0::int END AS male, CASE WHEN (titlefemale=1 OR titlemale=1 OR exclusivelymale=1 OR exclusivelyfemale=1) THEN 0::int ELSE 1::int END AS unknowngender, CASE WHEN (ceo=1 OR president=1) THEN 1::int ELSE 0::int END AS ceopres, CASE WHEN (chairman=1 OR ceo=1 OR cfo=1 OR coo=1 OR cio=1 OR cto=1 OR otherclvl=1 OR president=1) THEN 1::int ELSE 0::int END AS CLevel, CASE WHEN (chairman=1 OR ceo=1 OR cfo=1 OR coo=1 OR cio=1 OR cto=1 OR otherclvl=1 OR president=1 OR director=1 OR boardmember=1) THEN 1::int ELSE 0::int END AS board, CASE WHEN (chairman=1 OR ceo=1 OR cfo=1 OR coo=1 OR cio=1 OR cto=1 OR otherclvl=1 OR president=1 OR director=1 OR boardmember=1 OR vp=1 OR founder=1) THEN 1::int ELSE 0::int END AS vpandabove FROM copeoplebase LEFT JOIN namegender ON namegender.firstname=copeoplebase.firstname WHERE hasperson=1; --177547 The next table only keeps executive listings that have a valid portco primary key associated with them.   DROP TABLE CoPeopleKey; CREATE TABLE CoPeopleKey AS SELECT A.* FROM CoPeopleFull AS A JOIN (SELECT firstname, lastname, cname, datefirstinv, statecode, count(*) FROM CoPeopleFull WHERE firstname IS NOT NULL AND lastname IS NOT NULL AND cname IS NOT NULL AND datefirstinv IS NOT NULL AND statecode IS NOT NULL GROUP BY firstname, lastname, cname, datefirstinv, statecode HAVING COUNT(*)=1) AS B ON A.firstname=B.firstname AND A.lastname=B.lastname AND A.datefirstinv=B.datefirstinv AND A.cname=B.cname AND A.statecode=B.statecode; --176251 The next table identifies whether a person previously held executive positions.  CREATE TABLE CoPeopleSerial AS SELECT firstname, lastname, cname, datefirstinv, statecode, COALESCE(sum(hasperson),0) as prev, COALESCE(sum(ceo),0) as prevceo, COALESCE(sum(ceopres),0) as prevceopres, COALESCE(sum(founder),0) as prevfounder, COALESCE(sum(clevel),0) as prevclevel, COALESCE(sum(board),0) as prevboard, COALESCE(sum(vpandabove),0) as prevvpandabove, CASE WHEN sum(hasperson) >=1 THEN 1::int ELSE 0::int END AS serial, CASE WHEN sum(ceo) >=1 THEN 1::int ELSE 0::int END AS serialceo, CASE WHEN sum(ceopres) >=1 THEN 1::int ELSE 0::int END AS serialceopres, CASE WHEN sum(founder) >=1 THEN 1::int ELSE 0::int END AS serialfounder, CASE WHEN sum(clevel) >=1 THEN 1::int ELSE 0::int END AS serialclevel, CASE WHEN sum(board) >=1 THEN 1::int ELSE 0::int END AS serialboard, CASE WHEN sum(vpandabove) >=1 THEN 1::int ELSE 0::int END AS serialvpandabove FROM ( SELECT A.prefix, A.firstname, A.lastname, A.cname, A.datefirstinv, A.statecode, B.cname as prevcname, B.datefirstinv as prevdatefirstinv, B.statecode as prevstatecode, B.ceo, B.ceopres, B.founder, B.clevel, B.board, B.vpandabove, B.hasperson FROM CoPeopleKey AS A LEFT JOIN CoPeopleKey AS B ON A.firstname=B.firstname AND A.lastname=B.lastname AND A.datefirstinv > B.datefirstinv ) AS T GROUP BY firstname, lastname, cname, datefirstinv, statecode; --176251 The last table aggregates a ton of information on executives for each company. There is too much information to explain it all.   DROP TABLE copeopleagg; CREATE TABLE copeopleagg AS SELECT A.cname, A.datefirstinv, A.statecode, sum(hasperson) as numperson, sum(hastitle) as numtitled, CASE WHEN sum(ceopres) >=1 THEN 1::int ELSE 0::int END AS hasceopres, CASE WHEN sum(founder) >=1 THEN 1::int ELSE 0::int END AS hasfounder, CASE WHEN sum(clevel) >=1 THEN 1::int ELSE 0::int END AS hasclevel, CASE WHEN sum(board) >=1 THEN 1::int ELSE 0::int END AS hasboard, CASE WHEN sum(vpandabove) >=1 THEN 1::int ELSE 0::int END AS hasvpandabove, sum(female) as females, sum(male) as males, sum(unknowngender) as ugs, sum(doctor*female) as femaledoctors, sum(doctor*male) as maledoctors, sum(doctor*unknowngender) as ugdoctors, sum(ceopres*female) as femaleceos, sum(ceopres*male) as maleceos, sum(ceopres*unknowngender) as ugceos, sum(ceopres*female*doctor) as femaledoctorceos, sum(ceopres*male*doctor) as maledoctorceos, sum(ceopres*unknowngender*doctor) as ugdoctorceos, sum(founder*female) as femalefounders, sum(founder*male) as malefounders, sum(founder*unknowngender) as ugfounders, sum(founder*female*doctor) as femaledoctorfounders, sum(founder*male*doctor) as maledoctorfounders, sum(founder*unknowngender*doctor) as ugdoctorfounders, sum(clevel*female) as femaleclevels, sum(clevel*male) as maleclevels, sum(clevel*unknowngender) as ugclevels, sum(clevel*female*doctor) as femaledoctorclevels, sum(clevel*male*doctor) as maledoctorclevels, sum(clevel*unknowngender*doctor) as ugdoctorclevels, sum(board*female) as femaleboards, sum(board*male) as maleboards, sum(board*unknowngender) as ugboards, sum(board*female*doctor) as femaledoctorboards, sum(board*male*doctor) as maledoctorboards, sum(board*unknowngender*doctor) as ugdoctorboards, sum(vpandabove*female) as femaleabovevps, sum(vpandabove*male) as maleabovevps, sum(vpandabove*unknowngender) as ugabovevps, sum(vpandabove*female*doctor) as femaledoctorabovevps, sum(vpandabove*male*doctor) as maledoctorabovevps, sum(vpandabove*unknowngender*doctor) as ugdoctorabovevps, sum(prev*female) as femaleprevs, sum(prev*male) as maleprevs, sum(prev*unknowngender) as ugprevs, sum(prevceopres*female) as femaleprevceopres, sum(prevceopres*male) as maleprevceopres, sum(prevceopres*unknowngender) as ugprevceopres, sum(prevfounder*female) as femaleprevfounder, sum(prevfounder*male) as maleprevfounder, sum(prevfounder*unknowngender) as ugprevfounder, sum(prevclevel*female) as femaleprevclevel, sum(prevclevel*male) as maleprevclevel, sum(prevclevel*unknowngender) as ugprevclevel, sum(prevboard*female) as femaleprevboard, sum(prevboard*male) as maleprevboard, sum(prevboard*unknowngender) as ugprevboard, sum(prevvpandabove*female) as femaleprevvpandabove, sum(prevvpandabove*male) as maleprevvpandabove, sum(prevvpandabove*unknowngender) as ugprevvpandabove, sum(serial*female) as femaleserials, sum(serial*male) as maleserials, sum(serial*unknowngender) as ugserials, sum(serialceopres*female) as femaleserialceopres, sum(serialceopres*male) as maleserialceopres, sum(serialceopres*unknowngender) as ugserialceopres, sum(serialfounder*female) as femaleserialfounder, sum(serialfounder*male) as maleserialfounder, sum(serialfounder*unknowngender) as ugserialfounder, sum(serialclevel*female) as femaleserialclevel, sum(serialclevel*male) as maleserialclevel, sum(serialclevel*unknowngender) as ugserialclevel, sum(serialboard*female) as femaleserialboard, sum(serialboard*male) as maleserialboard, sum(serialboard*unknowngender) as ugserialboard, sum(serialvpandabove*female) as femaleserialvpandabove, sum(serialvpandabove*male) as maleserialvpandabove, sum(serialvpandabove*unknowngender) as ugserialvpandabove, sum(ceopres*serialceopres*female) as femaleceopresserialceopres, sum(ceopres*serialceopres*male) as maleceopresserialceopres, sum(ceopres*serialceopres*unknowngender) as ugceopresserialceopres, sum(founder*serialfounder*female) as femalefounderserialfounder, sum(founder*serialfounder*male) as malefounderserialfounder, sum(founder*serialfounder*unknowngender) as ugfounderserialfounder FROM CoPeoplekey AS A JOIN CoPeopleSerial AS B ON A.firstname=B.firstname AND A.lastname=B.lastname AND A.datefirstinv=B.datefirstinv AND A.cname=B.cname AND A.statecode=B.statecode GROUP BY A.cname, A.datefirstinv, A.statecode; --30413 Since this table is so big, it is a good idea to have a smaller, more manageable table to work with.  DROP TABLE copeopleaggsimple; CREATE TABLE copeopleaggsimple AS SELECT cname, datefirstinv, statecode, numperson, females, males, ugs, ugdoctors, maleserials+femaleserials+ugserials AS serials FROM copeopleagg; --30413 ===Fund People===Luckily, this process is much easier than the company people process. First we must simply load the data into the db.  DROP TABLE fundpeople; CREATE TABLE fundpeople( fundname varchar(255), fundyear int, prefix varchar(5), firstname varchar(50), lastname varchar(50), jobtitle varchar(150), prevpos varchar(255) );  \COPY fundpeople FROM 'Executives-Funds-NoFoot-normal.txt' WITH DELIMITER AS E'\t' HEADER NULL AS '' CSV --328994 The next table identifies degree and sex information about the executives of the fund.  DROP TABLE fundpeoplebase; CREATE TABLE fundpeoplebase AS SELECT fundpeople.*, CASE WHEN prefix='Ms' THEN 1::int WHEN prefix='Mr' THEN 0::int ELSE Null::int END AS titlefemale, CASE WHEN prefix='Ms' THEN 0::int WHEN prefix='Mr' THEN 1::int ELSE Null::int END AS titlemale, CASE WHEN prefix='Dr' THEN 1::int ELSE 0::int END AS doctor, CASE WHEN prefix IS NULL THEN 0::int ELSE 1::int END AS hastitle, CASE WHEN prefix IS NULL AND firstname IS NULL AND lastname IS NULL THEN 0::int ELSE 1::int END AS hasperson FROM fundpeople; --328994 The next table tries to identify the sex of the executive using the above defined namegender table. It only selects rows where a person is actually listed.  DROP TABLE FundPeopleFull; CREATE TABLE FundPeopleFull AS SELECT fundpeoplebase.*, exclusivelyfemale, exclusivelymale, CASE WHEN titlefemale=1 THEN 1::int WHEN exclusivelyfemale=1 AND exclusivelymale=0 AND (titlemale=0 OR titlemale IS NULL) THEN 1::int ELSE 0::int END AS female, CASE WHEN titlemale=1 THEN 1::int WHEN exclusivelymale=1 AND exclusivelyfemale=0 AND (titlefemale =0 OR titlefemale IS NULL) THEN 1::int ELSE 0::int END AS male, CASE WHEN (titlefemale=1 OR titlemale=1 OR exclusivelymale=1 OR exclusivelyfemale=1) THEN 0::int ELSE 1::int END AS unknowngender FROM fundpeoplebase LEFT JOIN namegender ON namegender.firstname=fundpeoplebase.firstname WHERE hasperson=1; --320915 The next table gives you information on executives aggregated by fund.  DROP TABLE FundPeopleAgg; CREATE TABLE FundPeopleAgg AS SELECT fundname, sum(female) as numfemale, sum(male) as nummale, sum(unknowngender) as numunknowngender, sum(doctor) as numdoctor, sum(female*doctor) as numfemaledoctor, sum(male*doctor) as nummaledoctor, sum(unknowngender*doctor) as numunknowngenderdoctor, sum(hastitle) as numtitled, sum(hasperson) as numpeople, CASE WHEN sum(hasperson) > 0 THEN sum(female)/sum(hasperson) ELSE NULL END as fracfemale, CASE WHEN sum(male) > 0 THEN sum(female)/sum(male) ELSE NULL END as ratiofemale FROM FundPeopleFull GROUP BY fundname; --21536 It is also good to have this information on firms. We do not pull firm people information from SDC. However, we have enough information to create it from preexisting tables.  DROP TABLE firmpeopleagg; CREATE TABLE firmpeopleagg AS SELECT _firmname as firmname, sum(numfemale) as firmwomen, sum(nummale) as firmmen, sum(numunknowngender) as firmugs, sum(numdoctor) as firmdoctors, sum(numpeople) as firmpeople, CASE WHEN sum(numpeople) > 0 THEN (sum(numfemale)/sum(numpeople))::real ELSE NULL END as firmfracwomen, CASE WHEN sum(nummale) > 0 THEN (sum(numfemale)/sum(nummale))::real ELSE NULL END as firmratiowomen FROM roundlineaggfunds AS A JOIN fundpeopleagg AS B ON A._fundname=B.fundname GROUP BY _firmname; --5273 ==Marcos's Code==This is code that a Rice student, Marcos Lee, wrote. I cleaned it and ran it. I have described the tables that I built and where they come from below. My code is located in: E:McNair\Projects\VentureXpert Database\vcdb3\LoadingScripts\MatchingEntrepsV3 If you have issues understanding my explanation, go to this location and read the query. Most of them are straight forward. ===Describing Stacks Created in Code=== CoPeopleBase: -Builds from copeople and titlelookup -Identifies what roles people played in their companies  namegender: -built from copeoplebase -identifies male/female/unknown  CoPeopleFull: -built from copeoplebase and namegender -builds more extensive information on executive including speficially what level of executive they are  CoPeopleKey: -built from CoPeopleFull -creates table where only executives with full primary keys are kept  CoPeopleSerial: -built from copeoplekey -keeps track of executives previous jobs at executive level  CoPoepleAgg: -built from copeoplekey and copeopleserial -gets extensive information on executives for each company  FundPeopleBae: -built from fundpeople -identifies male/female/doctor -hasperson column slightly weird because we can only have the lastname without prefix or first name and still have a 1 in column. Seems to be of little use/too broad  FundPeopleFull: -built from fundpeoplebase, namegender -adds in male/female   Fundpeopleagg: -built from fundpeoplefull -has aggregations of gender info for each fund  RoundLineJoinerLeanffWlistno: -built from rounlinejoinerleanff -adds listno to funds  RoundLineAggFunds: -built from roundlinejoinerleanffwlistno and rounlineaggfirms -if there are two funds from one firm that invest in same portco, we choose only one and leave the others behind  RoundLineAggWExit: -built from roundlineaggfirms, portcoexitupdated, roundlineaggfunds -adds in exit information for each company in roundlineaggfirms  FirmPerf: -built from roundlineaggwexit -adds in various performance measures for a given firm   PortCoFundDemo: -built from roundlinejoinerleanffclean and fundpeopleagg -gives information on executives of funds who invested in the portcos  PortCoPeopleMaster: -built from PortCoMaster, PortCoIndustry, PortCoPatent, PortCoSBIR, copeoplagg, PortCoFundDemo, CPI, statelookupint -huge amount of data about companies and their executives  RoundAggDistBase: -built from portcogeo, firmbogeo, roundlineaggwexit -creates geographic points using long, lat from geocoding  RoundAggDist: -Built from roundaggdistbase -gets actual distances between portcos and firms. if branch office exists and distance is less than distance to firm chooses that also generates random number  FirmPeopleAgg: -built from roundlineaggfunds, fundpeopleagg -finds information on executives from different firms  PortCoMatchmaster: -built from portcopatent, porcoindustry, portcosbir, copeopleaggsimple, portcoid -gets all information together about portcos  FirmMatchMaster: -built from firmperf, firmvars, firmpeopleagg, firmid -gets all information together about firms  RoundLineMasterBase: -built from portcomatchmaster, firmmatchmaster, roundaggdist, roundlineaggwexit -builds large amount of information about portcos and firms spceifically info about exits and distances  MatchMostNumerous: -built from roundlinemasterbase -finds max number of portcos invested in by a firm that also invested in the company grouping by  MatchHighestRandom: -built from matchmostnumerous -if two firms that invested in one company had the same number of max port cos this randomly chooses one company  FirmActiveYearsCode20: -built from roundlinejoinerleanffclean, porcoindustry -adds firmname to industry code not exactly sure why distinct is used in query  RealMatchesCode20: -built from MatchHighestRandom, PortCoIndustry -real matches between portcos and firms that invested in them including the code20  SyntheticFirmSetBaseCode20: -built from realmatchescode20, firmactiveyarscode20 -crossproduct of firms and portcos. finds firms that invested in same year as portco received first inv, firms invested in same type of company, and makes sure matches are unique  AllMatchKeys: -built from SyntheticFirmSetBaseCode20, RealMatchesCode20 -combines synthetic and real matches  SynthRoundAggDistBaseCode20: -built from allmatchkeys, portcogeo, firmbogeo -builds points for all portco, firm listings in allmatch keys  SynthRoundAddDistCode20: -built from synthroundaggdistvasecode20 -finds actual distance between portcos and firms using installed extensions chooses branch offices if distance between portco and bo less than firm  SynthFirmnameInduBlowoutCode20: -built from allmatchkeys, roundlinemasterbase -gets every firm combination and checks whehter the companies that those firms invested in are in the same general industry  SynthFirmNameroundInduHistCode20: -built from SynthFirmnameInduBlowoutcode20 -gets information by portco, firmname match about what the firms past investment patterns are  MasterWithSynthBaseCode20Portco: -built from Allmatchkeys, matchhighestrandom, synthroundaggdistcode20, sythnfirmnameroundinduhistcode20, synthfirmnameroundindutotalcode20, firmvars, copeopleaggsimple, portcomaster -builds a bunch of information about synthetic and real matches  SynthFirmnameRoundInduTotalCode20: -built from allmatchkeys, roundlinemasterbase -finds number of portcos in certain industries by firmnames  MasterWithSynthCode20Firms: -built with firmmatchmaster, allmatchkeys -matching a bunch of information to all firms  MasterWithSynthcode20: -built from masterwithsynthbasecode20portco, masterwithsynthcode20firms -gets a huge amount of info together on real and synthetic matches about firms and companies  MasterReals: -built from masterwithsynthcode20 -gets just real matches from code  MasterOneSynth: -built from masterwithsynthcode20 -gets just one randomly chosen synthetic match between companies and firms  MasterRealOneSynth: -built from masteronesynth, masterreals -combines the real and one synth table ==Ranking Tables and Graphs==This is a slight detour from the creation of VCDB3. However, this is a cool process because you actually get to use the data you've been working with. This process is extensive, but the queries are easy to understand. If you wish to have deeper understanding of the process, read the code. It is located in:  E:McNair\Projects\VentureXpert Database\vcdb3\LoadingScripts\RoundRanking.SQL First you must create a table that has aggregate round information grouped by cities and round year. Since this is a little difficult to picture, I will attach the code. DROP TABLE roundleveloutput; CREATE TABLE roundleveloutput AS SELECT city, statecode, roundyear AS year, SUM(rndamtestm*seedflag) AS seedamnt, SUM(rndamtestm*earlyflag) AS earlyamnt, SUM(rndamtestm*laterflag) AS lateramnt, SUM(rndamtestm*growthflag) AS selamnt, SUM(growthflag*dealflag) AS numseldeals FROM round GROUP BY city, statecode, roundyear; --30028 Next create a table that lists the all time SEL amount by city. Keep including the state code since this will ensure that you have the right city. City names are often repeated in different states. Next, create a table which lists unique city, state for every year since 1980. Then, build a table which matches portcos to the city, state, year blowout table for each year they were alive. This table should be relatively large since it lists companies once for every year they were alive up until the present. Then create a table that displays the number of companies alive in a city every year since 1980. Then add in a table that lists all of the information you have built in tables previously based on city, state, year. Also add in population. Then you can run the ranking queries. For states follow the same general process but group by states not cities and states.  If this explanation was not enough for you (it was not meant to be in depth) go to the location defined above and read the actual code. With the description I have given, you should be able to piece together what each query does. ==Master Tables==Throughout the creation of the database, there are inevitably some tables that are vital to create a solid foundation. The following tables are the master tables with a quick explanation:* '''Companybasecore'''- The base table for portcos. This is data that was drawn directly from SDC and was not changed other than for cleaning purposes. Count: 48001* '''BranchOfficeCore'''- The base table for branch offices. This is data drawn directly from SDC. Here only branch offices with distinct firm names are included. Count: 10032* '''FirmBaseCore'''- The base table for firms. This is also data taken directly from SDC and was not changed other than for cleaning purposes. Count: 15437* '''FundBaseCore'''- The base table for funds. This is also data taken directly from SDC and was not changed other than for cleaning purposes. Count: 28833* '''IPOCleanNoDups''' - This is the clean table of IPOs after being run through the matcher against portcos. It was cleaned manually and had duplicates removed. Count: 2136* '''IPONoDups'''- This is the table before the cleaning process of matching to portcos. There could be problems with this table as we used an aggregate function here. Be careful using this table. Count: 11149* '''MACleanNoDups'''- This is the clean table of MAs after being run through the matcher against portcos. It was cleaned manually and had duplicates removed. Count: 7171* '''MANoDups'''- This is the table before the cleaning process of matching to portcos. There could be problems with this table as we used an aggregate function here as well. Be careful using this table. Count: 119374* '''Round'''- This is the master round table. It has SEL flags attached to it and has the most round info. RoundBaseClean is also a decent table but has less information. This table is your best bet for round information. Count: 151323* '''RoundLineJoinerLeanFFClean'''- This is the master round table for joining purposes. It was cleaned and used for widespread joining purposes. Count: 163157* '''CoPeople'''- This is the base table for PortCo people information. It was pulled directly from SDC. Count: 194359* '''FirmBoGeo'''- This is the base table for firm/branch office geocoding. This table was cleaned and contains lat/long readings for firms and branch offices where the information was available. Count: 15437* '''PortCoGeo'''- This is the base table for portco geocoding. Table was cleaned and contains lat/long reading for portcos where the Google API returned a valid reading. Count: 48001* '''FirmPerf'''- This is a wide reaching table about the performance of firms. It was mainly used later in the project but is extremely useful. Count: 8336* '''FundPeople'''- This is the base table for fund people information. It was pulled directly from SDC. Count: 328994.* '''PortCoExitUpdated'''- This is the master exit table for portcos. The difference between this and PortCoExit is that Updated has two columns marking MAs and IPOs while the other has one column MAvsIPO. Use which ever one is more convenient. Count: 48001* '''PortCoMaster'''- This table is great. There's a ton of information on PortCos including SEL flags, round amounts, and industry classifications. Count: 48001