1、1Copyright 2007 Thomson Asia Pte. Ltd. All rights reserved.多元回归分析:渐进性n y = b0 + b1x1 + b2x2 + . . . bkxk + u2计量经济学导论计量经济学导论 刘愿刘愿渐进性的含义n如果误差并非正态分布,对任何的样本容量而言,t统计量、F统计量并非恰好服从t分布、F分布。n幸运的是,即使没有正态性假定,t统计量和F统计量仍然渐进的服从t分布、F分布,至少在大样本情况下使如此。3计量经济学导论计量经济学导论 刘愿刘愿一致性n在高斯-马尔科夫假定下,OLS估计是最优线性无偏的,但我们并非总能得到无偏的估计量。n
2、一致性是对一个估计量最起码的要求。在无法满足无偏性的情况下,我们可以搜集尽可能多的样本,即使n ,参数估计值的分布将逼近真实参数值。4计量经济学导论计量经济学导论 刘愿刘愿一致性的正式定义12nW,W0,0Wplim WnNnnnnY YYP WnWW n令是基于样本的参数 的估计值,则是 的一致估计量,对于任意一个正数,当否则,不是 的一致估计量。当是一致时,我们说 是的概率极限,记为:5计量经济学导论计量经济学导论 刘愿刘愿一致性的直观理解jjjjjnbbbbb如果估计量是一致的,那么随着样本容量的增加,的分布就越来越紧密地分布在的周围。当 趋向无穷时,的分布就紧缩成单一一个点。这意味着,
3、如果我们能够搜集到所需要的样本数据,我们就能让估计量任意接近于。6计量经济学导论计量经济学导论 刘愿刘愿当样本容量增加时的样本分布7计量经济学导论计量经济学导论 刘愿刘愿5.1OLSMLR.1MLR.4jjbb定理的一致性 在假定下,对所有的j=0,1, ,k,OLS估计量都是的一致估计。8计量经济学导论计量经济学导论 刘愿刘愿OLS的一致性n在高斯-马尔科夫假定下,OLS估计值是一致且无偏的。n类似的,我们可以像无偏性一样证明一致性,为此需要引入概率极限。9计量经济学导论计量经济学导论 刘愿刘愿简单回归中证明一致性211111110112110111111112112111111111111
4、10111plim,because ,0.iiiiiiiiiiiiiiiiiixx yxxxxxuxxxxxx xxx uxxnxx unxxCov x uVar xyxuand Cov x ubbbbbbbbbbb10计量经济学导论计量经济学导论 刘愿刘愿一个较弱的假定n为了得到无偏性,我们需要零条件均值假设E(u|x1, x2,xk) = 0n为了得到一致性,我们仅需要较弱的假定:零均值和零相关:E(u) = 0 ,Cov(xj,u) = 0, for j = 1, 2, , k.n不满足上述条件,OLS是有偏和不一致的。11计量经济学导论计量经济学导论 刘愿刘愿Deriving the
5、InconsistencynJust as we could derive the omitted variable bias earlier, now we want to think about the inconsistency, or asymptotic bias, in this case 1212112211022110, whereplim and, thatso , :You think :model TruexVarxxCovvxuuxyvxxybbbbbbbbb12计量经济学导论计量经济学导论 刘愿刘愿Asymptotic Bias (cont)nSo, thinking
6、 about the direction of the asymptotic bias is just like thinking about the direction of bias for an omitted variablenMain difference is that asymptotic bias uses the population variance and covariance, while bias uses the sample counterpartsnRemember, inconsistency is a large sample problem it does
7、nt go away as add data13计量经济学导论计量经济学导论 刘愿刘愿Summary of Direction of Asymptotic biasCorr(x1, x2) 0Corr(x1, x2) 0Positive asymptotic biasNegative asymptotic biasb2 0Negative asymptotic biasPositive asymptotic bias101014计量经济学导论计量经济学导论 刘愿刘愿Large Sample InferencenRecall that under the CLM assumptions, the
8、 sampling distributions are normal, so we could derive t and F distributions for testingnThis exact normality was due to assuming the population error distribution was normalnThis assumption of normal errors implied that the distribution of y, given the xs, was normal as well15计量经济学导论计量经济学导论 刘愿刘愿Lar
9、ge Sample Inference (cont)nEasy to come up with examples for which this exact normality assumption will failnAny clearly skewed variable, like wages, arrests, savings, etc. cant be normal, since a normal distribution is symmetricnNormality assumption not needed to conclude OLS is BLUE, only for infe
10、rence16计量经济学导论计量经济学导论 刘愿刘愿Central Limit TheoremBased on the central limit theorem, we can show that OLS estimators are asymptotically normalAsymptotic Normality implies that P(Zz)F(z) as n , or P(Zc, reject H26计量经济学导论计量经济学导论 刘愿刘愿27计量经济学导论计量经济学导论 刘愿刘愿nExample: Economic Model of CrimeNarr86为一个人被拘捕的次数;
11、Pcnv为以前被拘捕后被定罪的次数;Avgsen为过去定罪后被判刑的平均时间长度;Tottime为此人在年龄达到18岁后在1986年以前被送进监狱的总次数;Ptime86为1986年坐牢的月数;Qemp86为此人在1986年合法就业的季度数。0145868686narrpcnvptimeqemubbbb20123452,10%,2286860.0015,2725 0.00154.094.61,4.090.129uqxupcnvavgsentotimeptimeqemvRLMP x28计量经济学导论计量经济学导论 刘愿刘愿Asymptotic Efficiencyn Estimators bes
12、ides OLS will be consistentn However, under the Gauss-Markov assumptions, the OLS estimators will have the smallest asymptotic variancesn We say that OLS is asymptotically efficientn Important to remember our assumptions though, if not homoskedastic, not true29计量经济学导论计量经济学导论 刘愿刘愿The discussion in the simple regressiong(x) is any function of x, let zi=g(x), then u is uncorrelated with zi.30计量经济学导论计量经济学导论 刘愿刘愿