比较线性模型和Probit模型Logit模型

RevisedbyChenZhenin2021比较线性模型和Probit模型Logit模型研究生考试录取相关因素的实验 报告 软件系统测试报告下载sgs报告如何下载关于路面塌陷情况报告535n,sgs报告怎么下载竣工报告下载 研究目的通过对南开大学国际经济研究所1999级研究生考试分数及录取情况的研究，引入录取与未录取这一虚拟变量，比较线性概率模型与Probit模型，Logit模型，预测正确率。模型设定表1，南开大学国际经济研究所1999级研究生考试分数及录取情况见数据表obsYSCOREobsYSCOREobsYSCORE114013403326702752140135033268027331392360332690273413873703317002725138438033071026761379390328720266713784003287302638137841032874026191376420321750260101371430321760256111362440318770252121362450318780252131361460316790245140359470308800243150358480308810242161356490304820241170356500303830239180355510303840235190354520299850232200354530297860228210353540294870219220350550293880219230349560293890214240349570292900210250348580291910204260347590291920198270347600287930189280344610286940188290339620286950182300338630282960166310338640282970123320336650282330334660278定义变量SCORE：考生考试分数；Y：考生录取为1，未录取为0。上图为样本观测值。线性概率模型根据上面资料建立模型用Eviews得到回归结果如图：DependentVariable:YMethod:LeastSquaresDate:12/10/10Time:20:38Sample:197Includedobservations:97VariableCoefficientStd.Errort-StatisticProb.CSCORER-squaredMeandependentvarAdjustedR-squared.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodF-statisticDurbin-WatsonstatProb(F-statistic)参数估计结果为：+Se=(t=p=预测正确率：Forecast:YFActual:YForecastsample:197Includedobservations:97RootMeanSquaredErrorMeanAbsoluteErrorMeanAbsolutePercentageErrorTheilInequalityCoefficientBiasProportionVarianceProportionCovarianceProportion模型DependentVariable:YMethod:ML-BinaryLogit(Quadratichillclimbing)Date:12/10/10Time:21:38Sample:197Includedobservations:97Convergenceachievedafter11iterationsCovariancematrixcomputedusingsecondderivativesVariableCoefficientStd.Errorz-StatisticProb.CSCOREMeandependentvar.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.Restr.loglikelihoodAvg.loglikelihoodLRstatistic(1df)McFaddenR-squaredProbability(LRstat)ObswithDep=083Totalobs97ObswithDep=114得Logit模型估计结果如下pi=F(yi)=拐点坐标,其中Y=+预测正确率Forecast:YFActual:YForecastsample:197Includedobservations:97RootMeanSquaredErrorMeanAbsoluteErrorMeanAbsolutePercentageErrorTheilInequalityCoefficientBiasProportionVarianceProportionCovarianceProportion模型DependentVariable:YMethod:ML-BinaryProbit(Quadratichillclimbing)Date:12/10/10Time:21:40Sample:197Includedobservations:97Convergenceachievedafter11iterationsCovariancematrixcomputedusingsecondderivativesVariableCoefficientStd.Errorz-StatisticProb.CSCOREMeandependentvar.dependentvar.ofregressionAkaikeinfocriterionSumsquaredresidSchwarzcriterionLoglikelihoodHannan-Quinncriter.Restr.loglikelihoodAvg.loglikelihoodLRstatistic(1df)McFaddenR-squaredProbability(LRstat)ObswithDep=083Totalobs97ObswithDep=114Probit模型最终估计结果是pi=F(yi)=F+xi)拐点坐标,预测正确率Forecast:YFActual:YForecastsample:197Includedobservations:97RootMeanSquaredErrorMeanAbsoluteErrorMeanAbsolutePercentageErrorTheilInequalityCoefficientBiasProportionVarianceProportionCovarianceProportion预测正确率结论：线性概率模型RMSE=MAE=MAPE=Logit模型RMSE=MAE=MAPE=Probit模型RMSE=MAE=MAPE=由上面结果可知线性概率模型的RMSE、MAE、MAPE均远远大于Logit模型和Probit模型， 说明 关于失联党员情况说明岗位说明总经理岗位说明书会计岗位说明书行政主管岗位说明书 其误差率比Logit模型和Probit模型大很多，所以正确率远远小于Logit模型和Probit模型。而Logit模型和Probit模型的RMSE、MAE、MAPE相差很小，所以正确率相差不大。综上所诉，此数据可以用Logit模型和Probit模型代替线性概率模型进行 分析 定性数据统计分析pdf销售业绩分析模板建筑结构震害分析销售进度分析表京东商城竞争战略分析 。