版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
第一章简朴回归模型y=b0+b1x+u要求:1、一般最小二乘估计措施(OLS)2、OLS旳统计特征1ContentsWhatisthesimpleregressionmodel?Howtoderivetheordinaryleastsquares(OLS)estimates?PropertiesofOLSstatisticsandR2UnbiasednessofOLS
andVariancesoftheOLSestimators2Whatisthesimpleregressionmodel?y=b0+b1
x+u3SomeTerminologyInthesimplelinearregressionmodel,wherey=b0+b1x+u,wetypicallyrefertoyastheDependentVariable,orLeft-HandSideVariable,orExplainedVariable,orResponseVariable,orRegressand4SomeTerminology,cont.y=b0+b1x+u
Inthesimplelinearregressionofyonx,wetypicallyrefertoxastheIndependentVariable,orRight-HandSideVariable,orExplanatoryVariable,orRegressor,orCovariate,orControlVariables5ASimpleAssumption y=b0+b1x+uTheaveragevalueofu,theerrorterm,inthepopulationis0.Thatis,E(u)=0Thisisnotarestrictiveassumption,sincewecanalwaysuseb0
tonormalizeE(u)to0wage=b0+b1educ+u6ZeroConditionalMean y=b0+b1x+uWeneedtomakeacrucialassumptionabouthowuandxarerelatedWewantittobethecasethatknowingsomethingaboutxdoesnotgiveusanyinformationaboutu,sothattheyarecompletelyunrelated.Thatis,thatE(u|x)=E(u)=0,whichimpliesE(y|x)=b0+b1x,whichisoftencalled
PopulationRegressionFunction(PRF)7..x1x2E(y|x)asalinearfunctionofx,whereforanyx
thedistributionofyiscenteredaboutE(y|x)E(y|x)=b0+b1xyf(y)PopulationRegressionFunctionHowtoestimatetheparameters
b0andb1?8Howtoderivetheordinaryleastsquares(OLS)estimates?9OrdinaryLeastSquaresBasicideaofregressionistoestimatethepopulationparametersfromasampleLet{(xi,yi):i=1,…,n}denotearandomsampleofsizenfromthepopulationForeachobservationinthissample,itwillbethecasethat
yi=b0+b1xi+ui10....y4y1y2y3x1x2x3x4}}{{u1u2u3u4xyPopulationregressionline,sampledatapointsandtheassociatederrortermsE(y|x)=b0+b1x11DerivingOLSEstimatesToderivetheOLSestimatesweneedtorealizethatourmainassumptionofE(u|x)=E(u)=0alsoimpliesthatCov(x,u)=E(xu)=0Why?RememberfrombasicprobabilitythatCov(X,Y)=E(XY)–E(X)E(Y)12DerivingOLScontinuedWecanwriteour2restrictionsjustintermsofx,y,b0andb1,sincey=b0+b1x+u,u=y–b0–b1xE(y–b0–b1x)=0E[x(y–b0–b1x)]=0Thesearecalledmomentrestrictions13DerivingOLSusingM.O.M.ThemethodofmomentsapproachtoestimationimpliesimposingthepopulationmomentrestrictionsonthesamplemomentsWhatdoesthismean?RecallthatforE(X),themeanofapopulationdistribution,asampleestimatorofE(X)issimplythearithmeticmeanofthesampleSinxi/n14MoreDerivationofOLSWewanttochoosevaluesoftheparametersthatwillensurethatthesampleversionsofourmomentrestrictionsaretrueThesampleversionsareasfollows:15MoreDerivationofOLSGiventhedefinitionofasamplemean,andpropertiesofsummation,wecanrewritethefirstconditionasfollows16MoreDerivationofOLS17SotheOLSestimatedslopeis18SummaryofOLSslopeestimateTheslopeestimateisthesamplecovariancebetweenxandydividedbythesamplevarianceofxIfxandyarepositivelycorrelated,theslopewillbepositiveIfxandyarenegativelycorrelated,theslopewillbenegativeOnlyneedxtovaryinoursample19Whyordinaryleastsquares(OLS)?Intuitively,OLSisfittingalinethroughthesamplepointssuchthatthesumofsquaredresidualsisassmallaspossible,hencethetermleastsquaresTheresidual,û,isanestimateoftheerrorterm,u,andisthedifferencebetweenthefittedline(sampleregressionfunction)andthesamplepoint20.y4y1y2y3xySampleregressionline,sampledatapointsandtheassociatedestimatederrortermsE(y|x)=b0+b1x}{.x1x2x3x4}û2û3û4.û1•{21AlternateapproachtoderivationGiventheintuitiveideaoffittingaline,wecansetupaformalminimizationproblemThatis,wewanttochooseourparameterssuchthatweminimizethefollowing:22Alternateapproach,continuedIfoneusescalculustosolvetheminimizationproblemforthetwoparametersyouobtainthefollowingfirstorderconditions,whicharethesameasweobtainedbefore,multipliedbyn23AnSimpleExample(PRtextbook)(1)y(grade-pointaverage)(2)x(incomeofparentsin$1,000)(3)xi-x(4)yi-y(5)(xi-x)(yi-y)(6)(xi-x)
24.021.07.51.07.556.253.015.01.5.0.02.253.515.01.5.5.752.252.09.0-4.5-1.04.520.253.012.0-1.5.0.02.253.518.04.5.52.2520.252.56.0-7.5-.53.7556.252.512.0-1.5-.5.752.25y=3.0x=13.5S(xi-x)=0S(yi-y)=0S(xi-x)(yi-y)=19.50S(xi-x)2=162.00b1=0.120b0=y-b1x=1.37524PropertiesofOLSstatisticsandR225AlgebraicPropertiesofOLSThesumoftheOLSresidualsiszeroThus,thesampleaverageoftheOLSresidualsiszeroaswellThesamplecovariancebetweentheregressorsandtheOLSresidualsiszeroTheOLSregressionlinealwaysgoesthroughthemeanofthesample26AlgebraicProperties(precise)27Moreterminology28ProofthatSST=SSE+SSR29Goodness-of-FitHowdowethinkabouthowwelloursampleregressionlinefitsoursampledata?Cancomputethefractionofthetotalsumofsquares(SST)thatisexplainedbythemodel,callthistheR-squaredofregressionR2=SSE/SST=1–SSR/SSTWhere,0≤R2≤1....ûixy30UsingStataforOLSregressionsNowthatwe’vederivedtheformulaforcalculatingtheOLSestimatesofourparameters,you’llbehappytoknowyoudon’thavetocomputethembyhandRegressionsinStataareverysimple,toruntheregressionofyonx,justtyperegyx31UnbiasednessofOLSandVariancesoftheOLSestimators32UnbiasednessofOLSUnbiasedness:Assumethepopulationmodelislinearinparametersasy=b0+b1x+uAssumewecanusearandomsampleofsizen,{(xi,yi):i=1,2,…,n},fromthepopulationmodel.Thuswecanwritethesamplemodelyi=b0+b1xi+ui
AssumeE(u|x)=0andthusE(ui|xi)=0
AssumethereisvariationinthexiRandomsamplemeansCov(ui,uj)=033UnbiasednessofOLS(cont)Inordertothinkaboutunbiasedness,weneedtorewriteourestimatorintermsofthepopulationparameterStartwiththeformula34UnbiasednessofOLS(cont)35UnbiasednessofOLS(cont)Then,36UnbiasednessSummary
TheOLSestimatesofb1andb0areunbiasedProofofunbiasednessdependsonour4assumptions–ifanyassumptionfails,thenOLSisnotnecessarilyunbiasedRememberunbiasednessisadescriptionoftheestimator–inagivensamplewemaybe“near”or“far”fromthetrueparameter37VarianceoftheOLSEstimators
NowweknowthatthesamplingdistributionofourestimateiscenteredaroundthetrueparameterWanttothinkabouthowspreadoutthisdistributionisMucheasiertothinkaboutthisvarianceunderanadditionalassumption,soAssumeVar(u|x)=s2(Homoskedasticity)38VarianceofOLS(cont)Var(u|x)=E(u2|x)-[E(u|x)]2E(u|x)=0,sos2
=E(u2|x)=E(u2)=Var(u)Thuss2isalsotheunconditionalvariance,calledtheerrorvariances,thesquarerootoftheerrorvarianceiscalledthestandarddeviationoftheerrorCansay:E(y|x)=b0+b1xandVar(y|x)=s2Mayintroducethepropertiesofvariance,var(cx)=c2var(x),Var(c+x)=var(x)39..x1x2HomoskedasticCaseE(y|x)=b0+b1xyf(y|x)40.x
x1x2yf(y|x)HeteroskedasticCasex3..E(y|x)=b0+b1x41VarianceofOLS(cont)42VarianceofOLSSummaryThelargertheerrorvariance,s2,thelargerthevarianceoftheslopeestimat
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 畜禽粪污资源化处理利用技术方案
- 《储能电站电池性能故障排查报告模板》
- 商业综合体项目节能评估报告
- 热电联产项目绩效评价
- 包装PET瓶坯注塑异常处置方案
- 《露天矿作业人员职业健康管理制度》
- 中小企业绩效管理现存问题研究
- 家族式企业治理模式优化改进研究
- 部编版五年级上册示儿略读教学设计
- 2025年摄影技术(曝光控制技巧)试题及答案
- 2026年心理健康全科专任小学教师招聘考试笔试试题(含答案)
- GB/T 46585-2025建筑用绝热制品试件线性尺寸的测量
- 立体库安全管理制度
- 八年级英语下学期期末考试(深圳专用)(解析版)
- JTGT 3832-2018 公路工程预算定额 说明部分
- GB/T 44034-2024铁矿石矿浆的取样方法
- TCUWA40055-2023排水管道工程自密实回填材料应用技术规程
- 严重创伤患者紧急救治血液保障模式与输血策略中国专家共识(2024版)
- 第二单元6-10的认识和加减法(单元测试)-2024-2025学年人教版一年级上册数学
- 2021众海H6320火灾报警控制器(联动型)
- 健康产品营销策划方案(2篇)
评论
0/150
提交评论