2U24Workingwithdatesandtimes - PDF document

1 2[U]24Workingwithdatesandtimes Statareco
2[U]24Workingwithdatesandtimes Statarecordspuretimevalues(clocktimesindependentofdate)thesameway.Ratherthanthinkingofthenumericvalueasthenumberofmillisecondssince01jan1960,however,thinkofitasthenumberofmillisecondssincethebeginningoftheday.Forinstance,at2p.m.everyday,theairplanetakesofffromHoustonforLondon.Thenumericvalueassociatedwith2p.m.is50,400,000becausetherearethatmanymillisecondsbetweenthebeginningoftheday(00:00:00.000)and2p.m.Theadvantageofthinkingthiswayisthatyoucanadddatesandtimes.Whatisthedatetimevalueforwhentheplanetakesoffon21nov2006?Well,21nov200600:00:00.000is1,479,686,400,000(ignoringleapseconds),and1,479,686,400,000+50,400,000is1,479,736,800,000.Subtractingdatetimevaluesisuseful,too.Howmanyhoursaretherebetween21jan19527:23a.m.and21nov20063:14p.m.?Answer:(1,479,741,240,000�(�250,706,220,000))=3,600,000=480,679.85hours.Variablesthatrecordthenumberofmillisecondssince01jan1960andignoreleaps

2 econdsarecalled%tcvariables.Variablestha
econdsarecalled%tcvariables.Variablesthatrecordthenumberofmillisecondssince01jan1960andaccountforleapsecondsarecalled%tCvariables.Statahassevenotherkindsof%tvariables.Inmanyapplications,calendardatesbythemselvesaresufcient.Theapplicantwashiredon15jan2006,forinstance.Youcouldusea%tcvariabletorecordthatvalue,assigningsomearbitrarytimethatyouwouldignore,butitisbetterandeasiertousea%tdvariable.In%tdvariables,0stillcorrespondsto01jan1960,butaunitchangenowrepresentsanentiredayratherthanamillisecond.Thevalue1represents02jan1960.Thevalue�1represents31dec1959.Whenyousubtract%tdvariables,youobtainthenumberofdaysbetweendates.Inanancialapplication,youmightuse%tqvariables.In%tq,0representstherstquarterof1960,1representsthesecondquarter,and�1representsthelastquarterof1959.Whenyousubtract%tqvariables,youobtainthenumberofquartersbetweendates.Stataunderstandsnine%tformats:FormatBaseUnitsComment %tc01jan1960m

3 illisecondsignoresleapseconds%tC01jan196
illisecondsignoresleapseconds%tC01jan1960millisecondsaccountsforleapseconds%td01jan1960dayscalendardateformat%tw1960-w1weeks52ndweekmayhavemorethan7days%tmjan1960monthscalendarmonthformat%tq1960-q1quartersnancialquarter%th1960-h1half-years1half-year=2quarters%ty0A.Dyear1960meansyear1960%tb–daysuserdened Allformatsexcept%tyand%tbarebasedonthebeginningofJanuary1960.Thevalue0meanstherstmillisecond,day,week,month,quarter,orhalf-yearof1960,dependingonformat.Thevalue1isthemillisecond,day,week,month,quarter,orhalf-yearafterthat.Thevalue�1isthemillisecond,day,week,month,quarter,orhalf-yearbeforethat.Stata's%tyformatrecordsyearsasnumericvaluesanditcodesthemthenaturalway:ratherthan0meaning1960,1960means1960,andso2006alsomeans2006. 4[U]24Workingwithdatesandtimes .list namebdayx 1. Bill21Jan195222 2. May11Jul194818 3. Sam12Nov196025 4. Kay9Aug197516 Thedatalookne,butifwesetaboutusingthem,wewouldqu

4 icklydiscoverthereisnotmuchwecoulddowith
icklydiscoverthereisnotmuchwecoulddowithvariablebday.Variablebdaylookslikeadate,butitisjustastring.Weneedtoturnbdayintoa%tvariablethatStataunderstands:.genbirthday=date(bday,"DMY").list namebdayxbirthday 1. Bill21Jan195222-2902 2. May11Jul194818-4191 3. Sam12Nov196025316 4. Kay9Aug1975165699 Newvariablebirthdayisa%tdvariable.Theproblemnowisthat,whereasthenewvariableisperfectlyunderstandabletoStata,itisnotunderstandabletous.Naturallyenough,a%tdvariableneedsa%tdformat:.formatbirthday%td.list namebdayxbirthday 1. Bill21Jan19522221jan1952 2. May11Jul19481811jul1948 3. Sam12Nov19602512nov1960 4. Kay9Aug19751609aug1975 Usingournew%tdvariable,wecancreateavariablerecordinghowoldeachofthesesubjectswason01jan2000:.genage2000=(td(1jan2000)-birthday)/365.25.list namebdayxbirthdayage2000 1. Bill21Jan19522221jan195247.94524 2. May11Jul19481811jul194851.47433 3. Sam12Nov19602512nov196039.13484 4. Kay9Aug19751609aug197524.39

5 699 td()isafunctionthatmakesiteasytotype
699 td()isafunctionthatmakesiteasytotype%tddates.Therearealsofunctionstc(),tC(),tw(),tm(),tq(),andth()fortheother%tformats;see[D]datetime. 6[U]24Workingwithdatesandtimes lookslikewithdefaultformatting.Ifwewantedtoseethenumericvaluesbetter,wecouldchangedttohavea%20.0gcformat.Wewouldthenseethattherstvalueis1,479,113,983,000,thesecond1,479,195,409,000,andsoon.Wewillnotdothat.Instead,wewillputa%tcformatonour%tcvariable:.formatdt%tc.listiddtaction iddtaction 1. 100114nov200608:59:4315 2. 100215nov200607:36:4915 3. 100315nov200609:21:0711 4. 100215nov200614:57:3616 5. 100516nov200608:22:5312 6. 100116nov200608:36:4416 Variabledtisavariablewecanuse.Saywewantedtoknowhowmanyhoursithadbeensincethepreviousaction:.sortdt.genhours=hours(dt-dt[_n-1])(1missingvaluegenerated).formathours%9.2f.listiddtactionhours iddtactionhours 1. 100114nov200608:59:4315. 2. 100215nov200607:36:491522.62 3. 100315nov200609:21:07111.74 4.

6 100215nov200614:57:36165.61 5. 100516nov
100215nov200614:57:36165.61 5. 100516nov200608:22:531217.42 6. 100116nov200608:36:44160.23 Wesubtractedthepreviousvalueofdtfromdt,whichresultsinthenumberofmilliseconds.Convertingmillisecondstohoursiseasyenough:wejusthavetodivideby60601,000=3,600,000.Itiseasytoforgetormistypethatconstant,soweusedStata'shours()function,whichconvertsmillisecondstohours.hours(),andotherusefulfunctions,aredocumentedin[D]datetime.24.3DisplayingdatesandtimesA%tdvariableshouldhavea%tdformat,a%tcvariableshouldhavea%tcformat,andsoonforalltheother%tvariable–formatpairs.Youdothatbytypingformatvarname%td,formatvarname%tc,etc.Formats%tc,%tC,%td,%tw,%tm,%tq,%th,and%tyarecalledthedefault%tformats.Byspecifyingcodesfollowingthem,youcancontrolhowthevariableistobedisplayed.Inthepreviousexample,westartedwithastringvariablethatcontainedatimestampandlookedlike“TueNov1408:59:43CST2006”.Aftercreatinga%tcvariablefromit,andputtin

7 gthedefault%tcformatonit,ourdatetimesloo
gthedefault%tcformatonit,ourdatetimeslookedlike“14nov200608:59:43”.Belowwespecifya%tcformatthatmakesournewvariablelookjustliketheoriginal: 8[U]24Workingwithdatesandtimes 24.7BusinessdatesandcalendarsInadditiontothebuilt-indatetypesabove,suchas%tcand%td,Stataprovidesatypeyoucandene,denotedas%tbandcalledbusinessdatesandcalendars.Abusinesscalendarislikeanordinarycalendarwithsomedatescrossedout.Thecrossed-outdatescorrespondtothedatesonwhichthebusinessisclosed:November2011SuMoTuWeThFrSa 1234XX7891011XX141516171819X212223X25XX282930 Withrespecttoabusinessdate,yesterdayisthelastdaythebusinesswasopen,andtomorrowisthenextdaythebusinesswillbeopen.Considerdate=25nov2011.Ifdateisaregular(%td)datevariable,yesterday=date�1=24nov2011tomorrow=date+1=26nov2011Ifdateisabusiness(%tb)datevariable,yesterday=date�1=23nov2011tomorrow=date+1=28nov2011ThatisimportantbecausevariablescontainingdatesareoftenusedwithS

8 tata'slagandleadoperators;see[U]13.9Time
tata'slagandleadoperators;see[U]13.9Time­seriesoperators.Ifvariabletrading dateisanordinary%tdvari-able,thenL.trading datereallyisyesterdayandF.trading datereallyistomorrow.Butiftrading datehasanappropriatelydened%tbformat,L.trading dateistheprevioustradingdateandF.trading dateisthenexttradingdate.Youcanusebcalcreatetocreateabusinesscalendarbasedonthecurrentdataset.Alternatively,youcancreatealenamedcalname.stbcal,suchasnyse.stbcal.Afterthat,Statadeeplyunderstandsthenewformat%tbnyse.Formoreinformation,see[D]datetimebusinesscalendars.Businessdatesworkjustlikeregulardates,itisjustthatsomedatesarecrossedout.24.8ReferencesCox,N.J.2006.SpeakingStata:Timeofday.StataJournal6:124–137. .2010.Statatip68:Weekassumptions.StataJournal10:682–685. .2012.Statatip111:Moreonworkingwithweeks.StataJournal12:565–569.Samuels,S.J.,andN.J.Cox.2012.Statatip105:Dailydateswithmissingdays.StataJournal12:159–1