Lecture3_Models for Longitudinal_Panel Data and Mixed Models

《Lecture3_Models for Longitudinal_Panel Data and Mixed Models》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、ModelsforLongitudinal/PanelDataandMixedModelsChristopherFBaumBostonCollegeandDIWBerlinUniversityofAdelaide,June2010ChristopherFBaum(BC/DIW)Panel&MixedModelsAdelaide,June20101/88PaneldataanalysisFormsofpaneldataFormsofpaneldataTodefinetheproblemsofdatamanagement,considerada

2、tasetinwhicheachvariablecontainsinformationonNpanelunits,eachwithTtime-seriesobservations.Theseconddimensionofpaneldataneednotbecalendartime,butmanyestimationtechniquesassumethatitcanbetreatedassuch,sothatoperationssuchasfirstdifferencingmakesense.Thesedatamaybecommonlysto

3、redineitherthelongformorthewideform,inStataparlance.Inthelongform,eachobservationhasbothaniandtsubscript.ChristopherFBaum(BC/DIW)Panel&MixedModelsAdelaide,June20102/88PaneldataanalysisFormsofpaneldataLongformdata:.list,noobssepby(state)stateyearpopCT19903291967CT199533241

4、44CT20003411750MA19906022639MA19956141445MA20006362076RI19901005995RI19951017002RI20001050664ChristopherFBaum(BC/DIW)Panel&MixedModelsAdelaide,June20103/88PaneldataanalysisFormsofpaneldataHowever,youoftenencounterdatainthewideform,inwhichdifferentvariables(orcolumnsofthed

5、atamatrix)refertodifferenttimeperiods.Wideformdata:.list,noobsstatepop1990pop1995pop2000CT329196733241443411750MA602263961414456362076RI100599510170021050664Inavariantonthistheme,thewideformdatacouldalsoindextheobservationsbythetimeperiod,andhavethesamemeasurementfordiffe

6、rentunitsstoredindifferentvariables.ChristopherFBaum(BC/DIW)Panel&MixedModelsAdelaide,June20104/88PaneldataanalysisFormsofpaneldataTheformerkindofwide-formdata,wheretimeperiodsarearrayedacrossthecolumns,isoftenfoundinspreadsheetsoron-linedatasources.Theseexamplesillustrat

7、eabalancedpanel,whereeachunitisrepresentedineachtimeperiod.Thatisoftennotavailable,asdifferentunitsmayenterandleavethesampleindifferentperiods(companiesmaystartoperationsorliquidate,householdmembersmaydie,etc.)Inthosecases,wemustdealwithunbalancedpanels.Stata’sdatatransfo

8、rmationcommandsareuniquelyhandyinthatcontext.ChristopherFBaum(BC/DIW)Panel&MixedModelsAdelaide,J