The Stata News bull Executive Editor Karen Strope bull Production Supervisor Annette Fett

NEW

Stata 13 ships June 24 Order now at statacom

Treatment effectsbull Inverseprobabilityweights(IPW)bull Regressionadjustmentbull Propensity-scorematchingbull Covariatematchingbull Doublyrobustmethodsbull Continuousbinaryand

countoutcomes

Multilevelmixed modelsbull Negativebinomialbull Orderedlogisticbull Orderedprobitbull Multinomiallogisticbull GLMbull Hierarchicalmodels

Long stringsbull 2billioncharactersbull Textstringsbull Binarylargeobjects(BLOBs)bull ImportExportODBCSQLbull WorkjustlikeStatastrings

Generalized SEMbull Generalizedlinearresponses

binarycountorderedoutcomesbull Multilevelhierarchicalmodels

nestedandcrossedmodelsbull Randomslopesandinterceptsbull Fast

Power and sample sizebull Meansproportionsvariances

correlationsbull Casendashcontrolandcohortstudiesbull Interactivecontrolpanelbull Tabularresultsbull Automaticgraphs

Forecastingbull Timeseriesandpanelsbull Onetothousandsofequationsbull Identitiesbull Addfactorsbull Dynamicandstaticbull CIsviastochasticsimulationbull Comparescenarios

Effect sizesbull Meansbull ANOVAbull Linearregressionbull Confidenceintervalsbull CohenrsquosdHedgesrsquosg

GlassrsquosΔη2ω2

Panel databull Orderedoutcomesbull REorderedprobitbull REorderedlogisticbull ClusterndashrobustSEstorelax

distributionalassumptionsandallowforcorrelateddata

Project Managerbull Organizefiles(1-10000)bull Multipleprojectsbull Filteronfilenamesbull Clicktoopenbull Clicktorun

More statisticsbull Orderedprobitwithselectionbull Poissonwithendogenouscovariatesbull RobustSEsforquantileregressionbull MLestimationwithoutprogrammingbull Fractional-polynomialprefix

More documentationbull Treatment-effectsmanualbull Multilevelmixed-effectsmanualbull Powerandsample-sizemanualbull 2000morepagesbull 11000+totalpages

And morebull Factorvariablesshowlabelsbull Importdelimitedwithpreviewbull ImportfromHaverAnalyticsbull Businesscalendarsfromdatabull JavapluginAPIbull FTPandsecureHTTP

Highlights of Whatrsquos New in Stata 13

CO NFE RE N CE

Turn the page to learn more

Announcing

Stata 13 ships June 24 Order now at statacom

Long stringsbullEachupto2billioncharactersinlength(previousmaximumwas244)

bullAllstringfunctionsworkwiththem

bullAllofStataworkswiththem

bullReadwritefilesintofromthem

bullPlaintextandbinaryincludingBLOBsmdashbinarylargeobjects

Multilevel mixed-effects GLM Newestimators

bull Probitbull Complementarylogndashlogbull Orderedlogisticbull Orderedprobitbull Negativebinomialbull Multinomiallogistic

Newfeaturesbull Constraintsonvariancecomponentsbull Cluster-robustSEstorelaxdistributional

assumptionsandallowforcorrelateddatabull Posteriormodeandmeanestimatesof

randomeffects

Statanowfitsmultilevelgeneralizedlinearmodels

MultileveldatanaturallydividesintogroupsthathavesomethingincommonmdashstudentsattendingthesameschoolorpatientstreatedbythesamedoctorwithdoctorsthemselvesworkingatthesamehospitalStatarsquosmultilevelestimatorssupporttwo-leveldatathree-leveldataandhigher-leveldata

Generalizedoutcomescanbecontinuousbinarycountorderedorcategorical

Seethevideooverviewatstatacomvideos13multilevel-mixed-effects

LearneverythingaboutmultilevelmodelinginStataintheall-new356-pagemanualatstatacommanuals13me

YoucanreadentirefilesintothemevenbinaryfilessuchasWorddocumentsandJPEGimages

Seethevideooverviewatstatacomvideos13long-strings

LearneverythingaboutusinglongstringsinStataintheUserrsquos Guideatstatacommanuals13u124

Modelsbull Nestedmodels(hierarchical)bull Crossedmodelsbull Randominterceptsbull Randomcoefficients(slopes)

Nearlyeverybodyhaswantedstringslongerthan244charactersStatarsquosoldlimitPeopleaskedfor5001000even2000charactersSowesettledon2billion

TheselongstringsworkjustlikestringsinStatahavealwaysworked

AllofStataunderstandsthemmdashstringfunctionscommandsODBCimportexportsortbyetcUsetheminanalysisUsethemindatamanagementUsetheminprogramming

2

Power and sample sizeSolvefor

bull Powerbull Samplesizebull Minimumdetectableeffectbull Effectsize

Comparisonsofbull Means(t-tests)bull Proportionsbull Variancesbull Correlations

Designsbull Matchedcasendashcontrolstudiesbull Cohortstudiesbull Cross-sectionalstudies

SayweareplanninganexperimenttodeterminewhetherstudentswhopreparefortheSATexamobtainhighermathscoresbytakingclassesratherthanstudyingindependentlyThenationalaveragemathscoreis520withastandarddeviationof135Wewanttoseethepowerobtainedforsamplesizesof100through500whenscoresincreaseby204060and80pointsorequivalentlywhenaveragescoresincreaseto540560580and600Wetype

power twomeans 520 (540 560 580 600) n(100 200 300 400 500) sd(135) graph

Weassumedabovethatthosestudyingindependentlywillobtainthenationalaverageof520andthatthestandarddeviationswillbe135forbothgroupsWecouldrelaxeitherassumptionwithpower

Wemightbemoreinterestedinthesamplesizenecessarytodetecttheincreasedscoresatpowerlevelsof08and09Inthatcasewetype

power twomeans 520 (540 560 580 600) power(08 09) sd(135) graph

Inourexamplewedealtwithacomparisonofmeansacrosstwoindependentsamplespowercanalsoproducecomparisonsofproportionsvariancesandcorrelationsforonesampletwosamplesorpairedsamples

Seethevideooverviewatstatacomvideos13power-and-sample-size

LearneverythingaboutpowerandsamplesizeinStataintheall-new225-pagemanualatstatacommanuals13pssandinparticularseetheintroductionatstatacommanuals13pss_intro

Tablesandgraphsbull Automatedbull Custom

3

Treatment effectsEstimators

bull Inverseprobabilityweights(IPW)bull Propensity-scorematchingbull Covariatematchingbull Regressionadjustmentbull Doublyrobustmethods

rsaquo AugmentedIPWrsaquo IPWwithregressionadjustment

Featuresbull Multilevelandmultivaluedtreatmentsbull Averagetreatmenteffects(ATEs)bull ATEsonthetreated(ATETs)bull Potential-outcomemeans(POMs)bull Continuousbinaryandcountoutcomes

Endogenoustreatmentestimatorsbull Linearregressionbull Poissonregression

Atreatmentisanewdrugregimenasurgicalprocedureatrainingprogramorevenanadcampaignintendedtoaffectanoutcomesuchasbloodpressuremobilityemploymentorsales

Inthebestofworldswewouldmeasurethedifferenceinoutcomesbydesigninganexperimentthatassignssubjectsrandomlytothetreatmentandcontrol

WecanrsquotalwaysdothatWhenweneedtomakedowithobservationaldatamdashwhenthesubjectsthemselveschoosewhethertobetreatedorthechoiceisotherwisenonrandommdashweneedatreatment-effectsestimator

SaywehaveatrainingprograminwhichparticipantsenrollvoluntarilyIntherawdataparticipantsdopoorlyrelativetononparticipantsevenafterthetrainingEvensothetrainingprogrammighthaveimprovedtheiroutcomesmdashsayhourlywagesmdashoverwhattheywouldhavebeenTodeterminethisweneedatreatment-effectsestimator

Therearedifferenttreatment-effectestimatorsfordifferentsituationsLetustellyouaboutthem

WhenweknowthedeterminantsofparticipationtheappropriateestimatorsincludeIPWandpropensity-scorematchingWemighttype

teffects ipw (wage) (trained x1 x2)

teffects psmatch (wage) (trained x1 x2)

WhenweinsteadknowthedeterminantsofoutcometheappropriateestimatorsincluderegressionadjustmentandcovariatematchingWemighttype

teffects ra (wage x1 x3) (trained)

teffects nnmatch (wage x1 x3) (trained)

WhenweknowbothwecanusethedoublyrobustestimatorsmdashaugmentedIPWandIPWwithregressionadjustment

Wemighttype

teffects aipw (wage x1 x3) (trained x1 x2)

teffects ipwra (wage x1 x3) (trained x1 x2)

Weonlyneedtoberightaboutoneoftheadjustmentsmdashwageneedstobeafunctionofx1andx3ortrainedneedstobeafunctionofx1andx2Thatisafeatureofthedoublyrobustmethods

4

ToobtainthedoublyrobustIPWregression-adjustedresultswetype

teffects ipwra (wage x1 x3) (trained x1 x2)

Iteration 0 EE criterion = 2523e-16Iteration 1 EE criterion = 4680e-30

Treatment-effects estimation Number of obs = 1000Estimator IPW regression adjustmentOutcome model linearTreatment model logit------------------------------------------------------------------------------ | Robust wage | Coef Std Err z Pgt|z| [95 Conf Interval]-------------+----------------------------------------------------------------ATE | trained | (1 vs 0) | 4646677 080218 579 0000 3074432 6218921-------------+----------------------------------------------------------------POmean | trained | 0 | 8275523 0628417 13169 0000 8152356 8398691------------------------------------------------------------------------------

Theoutputrevealsthattheaveragetreatmenteffect(ATE)mdashtheeffectwewouldhaveobservedhadtheentirepopulationbeentreatedmdashis046meaning46centsmoreinthehourlywage(Thebaselinewageis$828Thisisthewagehadnoonebeentreated)

Ifweinsteadwanttheaveragetreatmenteffectforthetreated(ATET)weaddtheatetoption

teffects ipwra (wage x1 x3) (trained x1 x2) atet

Iteration 0 EE criterion = 2523e-16Iteration 1 EE criterion = 3211e-30

Treatment-effects estimation Number of obs = 1000Estimator IPW regression adjustmentOutcome model linearTreatment model logit------------------------------------------------------------------------------ | Robust wage | Coef Std Err z Pgt|z| [95 Conf Interval]-------------+----------------------------------------------------------------ATET | trained | (1 vs 0) | 8212578 0948967 865 0000 6352636 1007252-------------+----------------------------------------------------------------POmean | trained | 0 | 7974128 0886919 8991 0000 7800295 814796------------------------------------------------------------------------------

TheATETis82centsmoreperhouroverthebaselineamongthetreatedwhichis$797

Formoreinformationontreatmenteffectsseethevideooverviewatstatacomvideos13treatment-effects

LearneverythingabouttreatmenteffectsinStataintheall-new158-pagemanualatstatacommanuals13te

5

ToobtainthedoublyrobustIPWregression-adjustedresultswetype

teffects ipwra (wage x1 x3) (trained x1 x2)

Iteration 0 EE criterion = 2523e-16Iteration 1 EE criterion = 4680e-30

Treatment-effects estimation Number of obs = 1000Estimator IPW regression adjustmentOutcome model linearTreatment model logit------------------------------------------------------------------------------ | Robust wage | Coef Std Err z Pgt|z| [95 Conf Interval]-------------+----------------------------------------------------------------ATE | trained | (1 vs 0) | 4646677 080218 579 0000 3074432 6218921-------------+----------------------------------------------------------------POmean | trained | 0 | 8275523 0628417 13169 0000 8152356 8398691------------------------------------------------------------------------------

Theoutputrevealsthattheaveragetreatmenteffect(ATE)mdashtheeffectwewouldhaveobservedhadtheentirepopulationbeentreatedmdashis046meaning46centsmoreinthehourlywage(Thebaselinewageis$828Thisisthewagehadnoonebeentreated)

Ifweinsteadwanttheaveragetreatmenteffectforthetreated(ATET)weaddtheatetoption

teffects ipwra (wage x1 x3) (trained x1 x2) atet

Iteration 0 EE criterion = 2523e-16Iteration 1 EE criterion = 3211e-30

Treatment-effects estimation Number of obs = 1000Estimator IPW regression adjustmentOutcome model linearTreatment model logit------------------------------------------------------------------------------ | Robust wage | Coef Std Err z Pgt|z| [95 Conf Interval]-------------+----------------------------------------------------------------ATET | trained | (1 vs 0) | 8212578 0948967 865 0000 6352636 1007252-------------+----------------------------------------------------------------POmean | trained | 0 | 7974128 0886919 8991 0000 7800295 814796------------------------------------------------------------------------------

TheATETis82centsmoreperhouroverthebaselineamongthetreatedwhichis$797

Formoreinformationontreatmenteffectsseethevideooverviewatstatacomvideos13treatment-effects

LearneverythingabouttreatmenteffectsinStataintheall-new158-pagemanualatstatacommanuals13te

5

ForecastingbullTimeseriesandpaneldatasets

bullMultipleestimationresultsrsaquo OLSVARsVECsARIMA3SLSmorersaquo EstimatedwithStataorrsaquo Obtainedfromoutsidesourcesrsaquo Oneequationorthousands

bullIdentities

bullAddfactorsandotheradjustments

bullDynamicorstatic(1stepahead)forecasts

bullSolvesimultaneoussystems

bullConfidenceintervalsviastochasticsimulation

bullCompareforecastsofalternativescenarios

LetusshowyouWewillforecastasimpleseven-equationmodelproducedbyoneestimationcommandbutwecouldforecastamodelwiththousandsofequationsobtainedfromthousandsofestimationcommands

InoursimpleexamplewefitaclassicmodeloftheUSeconomytopre-WorldWarIIdatabyusing3SLS

reg3 (c p Lp w) (i p Lp Lk) (wp y Ly yr) endog(w p y) exog(t wg g)

estimates store klein

Nowwetellforecastaboutourestimationresultsandthemodelrsquosidentities

forecast create kleinmodelForecast model kleinmodel started

forecast estimates kleinAdded estimation results from reg3Forecast model kleinmodel now contains 3 endogenous variables

forecast identity y = c + i + gForecast model kleinmodel now contains 4 endogenous variables

forecast identity p = y - t - wpForecast model kleinmodel now contains 5 endogenous variables

forecast identity k = Lk + iForecast model kleinmodel now contains 6 endogenous variables

forecast identity w = wg + wpForecast model kleinmodel now contains 7 endogenous variables

forecast exogenous wg g t yrForecast model kleinmodel now contains 4 declared exogenous variables

6

EvenbetterwecouldhaveenteredthepreviousmodelintoforecastrsquosControlPanel

Wecouldgraphtheresults

Seethevideooverviewatstatacomvideos13forecast

Learneverythingaboutforecastbyreadingthemanualatstatacommanuals13forecast

7

EvenbetterwecouldhaveenteredthepreviousmodelintoforecastrsquosControlPanel

Wecouldgraphtheresults

Seethevideooverviewatstatacomvideos13forecast

Learneverythingaboutforecastbyreadingthemanualatstatacommanuals13forecast

7

Generalized SEM Generalizedlinearresponses

bull Continuousmdashlinearlog-gammabull Binarymdashprobitlogitcomplementarylog-logbull CountmdashPoissonnegativebinomialbull Categoricalmdashmultinomiallogitbull Orderedmdashorderedlogitorderedprobit

Multilevelhierarchicaldatabull Nested2levels3levelsmorelevelsbull Crossedbull Latentvariablesatdifferentlevelsbull Randominterceptsbull Randomslopes(paths)bull Mixedmodels

PathBuildersupportsGLMandmultilevelmodels

Newmodelsbull CFAwithbinarycountorordinalmeasurementsbull MultilevelCFAbull Multilevelmediationbull Item-responsetheory(IRT)bull Latentgrowthcurveswithrepeatedmeasurementsof

binarycountorordinalresponsesbull Selectionmodelsbull Endogenoustreatmenteffectsbull AnymultilevelSEMwithgeneralizedlinearresponses

SaywehaveatestdesignedtoassessmathematicalperformanceThedatarecordasetofbinaryvariablesmeasuringwhetherindividualanswerswerecorrectThetestwasadministeredtostudentsatvariousschools

Wepostulatethatperformanceonthequestionsisdeterminedbyunobserved(latent)mathematicalaptitudeandbyschoolqualityrepresentinglatentcharacteristicsoftheschool

InthediagramthevaluesofthelatentvariableSchQualareconstantwithinschoolandvaryacrossschools

Wecanfitthemodelfromthepathdiagrambypressing

Resultswillappearonthediagram

8

Orwecanskipthediagramandtypetheequivalentcommand

gsem ( MathApt -gt q1 q2 q3 q4) (SchQual[school] -gt q1 q2 q3 q4) logit

Eitherwaywegetthesameresults

Generalized structural equation model Number of obs = 2500Log likelihood = -65061646

( 1) [q1]SchQual[school] = 1 ( 2) [q2]MathApt = 1------------------------------------------------------------------------------------- | Coef Std Err z Pgt|z| [95 Conf Interval]--------------------+----------------------------------------------------------------q1 lt- | SchQual[school] | 1 (constrained) | MathApt | 1437913 1824425 788 0000 1080333 1795494 _cons | 0459474 1074647 043 0669 -1646795 2565743--------------------+----------------------------------------------------------------q2 lt- | SchQual[school] | 1522361 0823577 185 0065 -0091821 3136543 | MathApt | 1 (constrained) _cons | -377969 0518194 -729 0000 -4795332 -2764047--------------------+----------------------------------------------------------------q3 lt- | SchQual[school] | 5194866 0965557 538 0000 3302408 7087324 | MathApt | 8650544 1098663 787 0000 6497204 1080388 _cons | 026989 0667393 040 0686 -1038175 1577955--------------------+----------------------------------------------------------------q4 lt- | SchQual[school] | 6085149 119537 509 0000 3742266 8428032 | MathApt | 1721957 2466729 698 0000 1238487 2205427 _cons | -3225736 0845656 -381 0000 -4883191 -1568281--------------------+----------------------------------------------------------------var(SchQual[school])| 4167718 1222884 2344987 7407238 var(MathApt)| 1004914 1764607 7122945 1417744-------------------------------------------------------------------------------

MathaptitudehasalargervarianceandloadingsthanschoolqualityThusmathaptitudeismoreimportantthanschoolalthoughschoolisstillimportant

Seethevideooverviewatstatacomvideos13gsem

SeetheintroductiontogeneralizedSEMatstatacommanuals13semintro1orseetheentireSEMmanualatstatacommanuals13semInthemanualweadded18newworkedexamplesillustratingthemodelsyoucanfitwiththenewfeatures

9

Orwecanskipthediagramandtypetheequivalentcommand

gsem ( MathApt -gt q1 q2 q3 q4) (SchQual[school] -gt q1 q2 q3 q4) logit

Eitherwaywegetthesameresults

Generalized structural equation model Number of obs = 2500Log likelihood = -65061646

( 1) [q1]SchQual[school] = 1 ( 2) [q2]MathApt = 1------------------------------------------------------------------------------------- | Coef Std Err z Pgt|z| [95 Conf Interval]--------------------+----------------------------------------------------------------q1 lt- | SchQual[school] | 1 (constrained) | MathApt | 1437913 1824425 788 0000 1080333 1795494 _cons | 0459474 1074647 043 0669 -1646795 2565743--------------------+----------------------------------------------------------------q2 lt- | SchQual[school] | 1522361 0823577 185 0065 -0091821 3136543 | MathApt | 1 (constrained) _cons | -377969 0518194 -729 0000 -4795332 -2764047--------------------+----------------------------------------------------------------q3 lt- | SchQual[school] | 5194866 0965557 538 0000 3302408 7087324 | MathApt | 8650544 1098663 787 0000 6497204 1080388 _cons | 026989 0667393 040 0686 -1038175 1577955--------------------+----------------------------------------------------------------q4 lt- | SchQual[school] | 6085149 119537 509 0000 3742266 8428032 | MathApt | 1721957 2466729 698 0000 1238487 2205427 _cons | -3225736 0845656 -381 0000 -4883191 -1568281--------------------+----------------------------------------------------------------var(SchQual[school])| 4167718 1222884 2344987 7407238 var(MathApt)| 1004914 1764607 7122945 1417744-------------------------------------------------------------------------------

MathaptitudehasalargervarianceandloadingsthanschoolqualityThusmathaptitudeismoreimportantthanschoolalthoughschoolisstillimportant

Seethevideooverviewatstatacomvideos13gsem

SeetheintroductiontogeneralizedSEMatstatacommanuals13semintro1orseetheentireSEMmanualatstatacommanuals13semInthemanualweadded18newworkedexamplesillustratingthemodelsyoucanfitwiththenewfeatures

9

Project Manager

Effect sizes Comparisonofmeans

bull Cohenrsquosdbull Hedgesrsquosgbull GlassrsquosΔ

rsaquo Pointbiserialcorrelationrsaquo Estimatedfromdataorpublishedsummarystatistics

VarianceexplainedbyregressionandANOVAbull Eta-squaredandpartialeta-squared(η2)bull Omega-squaredandpartialomega-squared(ω2)bull Estimatedfromdatabull Overallstatisticsfromdataorpublishedsummarystatistics

Allwithconfidenceintervals

Effectsizesprovideresultsthewaymanyresearcherswanttoseethemandthewaymanyjournalsrequire

CohenrsquosdHedgesrsquosgandGlassrsquosΔhelptojudgepracticalasopposedtostatisticalsignificance

Eta-squaredandomega-squaredmeasurevarianceexplainedinANOVAandregressionmodelsbothoverallandterm-by-term

Seethevideooverviewatstatacomvideos13effect-sizes

LearneverythingabouteffectsizesinStatainthemanualentryatstatacommanuals13esize

bullCreategroupsintheprojecttocategorizefiles

bullPortableprojects

bullRelativeorabsolutepaths

bullOrganizedo-filesado-filesdatasetsrawfilesetc

bullManagehundredseventhousandsoffilesperproject

bullManagemultipleprojects

SeeallthefilesataglanceFilteronfilenamesClicktoopeneditorrun

Seethevideoatstatacomvideos13project-managerbutyoursquovegottotryit

10

And lots more Weshouldmentiontheimprovedhelp-filesearchingandPoissonwithendogenouscovariatesandthatStatanowsupportsFTPandsecureHTTPandfastPDFmanualnavigationandsevennoncentral-tandnoncentral-FfunctionsandJavaAPIandorderedprobitwithHeckman-stylesampleselectionandthenewwayofestimatingMLmodelswithoutwritinganevaluatorprogramandthenewfractional-polynomialprefixcommandandintraclasscorrelationsandthatquantileregressioncannowproducerobustestimatesofstandarderrorsandthatfactorvariablesnowsupportvaluelabelsforlabelingoutputandthenewwaytoimportdatafromHaverAnalyticsandautomaticbusiness-calendarcreationandthenewimportcommandsthatmakereadingdatareallyeasyandhowyoucancreateWordandExcelfilesfromStataandhowyoucansolvearbitrarynonlinearsystemsandalotofotherthings

AndStata13containsthemostrequestedfeatureofallYoucannowtypeclstocleartheResultsWindow

Learnmoreabouteverythingatstatacomstata13

StatTransfer Version 12bullSupportsnewStata13datasets

bullFullysupportsStatalongstringsandBLOBs

StatTransfer12alsoaddssupportforExcel2013gretlJMPVersion10JMPCompressedFilesSPSS21gsavCompressedFilesSAS64BitCatalogsandSYSTAT13

Italsoaddscross-platformlicensingOneactivationcodecanbeusedonWindowsMacorLinux

Findoutmoreatstatacomproductsstat-transfer

statacomgiftshop

Panel datamdashnew estimators and extensionsNewestimators

bull Random-effectsorderedprobit

bull Random-effectsorderedlogit

ItisdifficulttosaypaneldatawithoutsayingrandomeffectsPaneldataisrepeatedobservationsonindividualsRandomeffectsareindividual-leveleffectsthatareunrelatedtoeverythingelseinthemodel

Saywehavedataon4708employeesofalargemultinationalcorporationWehaverepeatedobservationsontheseemployeesovertheyearsOnaveragewehavesixyearsofdataForsomeemployeeswehavefifteenyears

Ourdataincludeprofessionalstatus(123or4)ageeducationandyearsofjobexperience

Wefitthismodel(showntotheright)

WefindthatprofessionalstatusincreaseswitheducationandexperienceWealsofindthatindividualshavealargepermanentcomponent(sigma2_uthevarianceoftherandomeffectisbothlargeandsignificant)

Seethevideooverviewatstatacomvideos13panel-data

Learnmoreaboutthenewestimatorsandextensionsinthemanualentriesatstatacommanuals13xtoprobitandstatacommanuals13xtologit

Clusterndashrobuststandarderrorsbull Relaxdistributionalassumptionsbull Allowforcorrelateddatabull Availableonnewestimatorsbull Alsoavailableonprobitlogitcomplementarylog-logandPoisson

11

StataCorp4905LakewayDriveCollegeStationTX77845-4512USA

Return service requested

Contact us979-696-4600 979-696-4601(fax)servicestatacom statacomPleaseincludeyourStataserialnumberwithallcorrespondence

Find a Stata distributor near you statacomworldwide

Copyright2013byStataCorpLPStataisaregisteredtrademarkofStataCorpLP

StataCorp Stata blogstatacomStataCorp +Stata

TheconferenceincludesinadditiontousercontributionspresentationsbyStataCorpdevelopersonnewStata13featuresMeetwiththedeveloperswhowroteStata13

DonrsquotmissthisopportunitytoexchangeideaswithStatausersanddevelopers

Seetheprogramandregisteronlineat

statacomnew-orleans13

Dates July 18ndash19 2013Venue Hyatt French Quarter New Orleans 800 Iberville Street New Orleans Louisiana Cost $195 regular $75 student $45 dinner at Tujaguersquos (optional)

Join us in the Big Easy and learn about Stata 13MixandminglewithStatausersanddevelopersAfteradayoflearningandnetworkingjoinusThursdayeveningforarelaxingauthenticcreoledinnerattherenownedTujaguersquosWhileintownwhynottakeinajazzbandtourtheFrenchQuarterorexploretheworld-famousBourbonStreet

Make New Orleans your destination

CO NFE RE N CE

Stata 13 ships June 24 Order now at statacom