在R（plm）中选择面板回归模型

Question

我正在尝试找出如何在R中使用面板回归，并且不确定是否要使用FE，RE，池化或模型之间

你好，

我是面板回归的新手，正努力使自己步入正轨。我目前正在使用一种模型来预测资金的年度总回报，其中包括连续，分类和互动期限。我的索引是资金ID和年份，分类变量是每个级别的虚拟变量（不包括1个虚拟对象）。由于我想包括时变变量（虚拟变量），因此我建立了具有effects =“ time”的固定效果模型，因为R不再删除这些变量，但已阅读到使用model =“ pooling”也是一种合适的方法使用时不变变量。我还建立了一个随机的“中间”模型，因为这些模型使我可以将它们保留在其中。“中间”模型的调整后R2很高，但是关于该模型估计器的含义似乎很少。这是否是正确的使用方法。基本上，我一直在尝试阅读所有我能找到的最佳方法，而且似乎有很多不同的方法，所以我的头有些微旋转！我对如何判断使用哪种最佳模型/是否真的误入歧途的任何见解将不胜感激。

固定效果模型（使用model =“ within”）

Europewithin <- plm(TotalReturnAnnual~ UKDummy+ GermanyDummy +  NetherlandsDummy + FranceDummy + ItalyDummy + FinlandDummy + Residential + IndustrialLogistics + Multi.Sector + Office + Retail +  CoreClosedEnd+ GAV +  I(GAV^2)+ Age + I(Age^2)+ gearingLag  + I(gearingLag^2)+ InterestRate +CPIGrowth + BBBBonds + InflationSurprise + M1MoneySupply  + InterestRate + Dummy2009Onwards+ GermanyDummy:Dummy2007_2008 + UKDummy:Dummy2009Onwards + UKDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2009Onwards + GermanyDummy:Dummy2007_2008 + GermanyDummy:Dummy2009Onwards + FranceDummy:Dummy2007_2008 + FranceDummy:Dummy2009Onwards + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2009Onwards + FinlandDummy:Dummy2007_2008 + FinlandDummy:Dummy2009Onwards + gearingLag:CoreDummy + I(gearingLag^2):CoreDummy + BrentCrude + RealGDPGrowth + CPIGrowth:InflationSurprise + RealGDPGrowth:InflationSurprise +gearingLag:UpmarketDummy + gearingLag:DownMarketDummy , data=panel, model="within", index = c("FundID", "Year"), effect="time")

合并模型

Europepool <- plm(TotalReturnAnnual~ UKDummy+ GermanyDummy +  NetherlandsDummy + FranceDummy + ItalyDummy + FinlandDummy + Residential + IndustrialLogistics + Multi.Sector + Office + Retail +  CoreClosedEnd+ GAV +  I(GAV^2)+ Age + I(Age^2)+ gearingLag  + I(gearingLag^2)+ InterestRate +CPIGrowth + BBBBonds + InflationSurprise + M1MoneySupply  + InterestRate + Dummy2009Onwards+ GermanyDummy:Dummy2007_2008 + UKDummy:Dummy2009Onwards + UKDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2009Onwards + GermanyDummy:Dummy2007_2008 + GermanyDummy:Dummy2009Onwards + FranceDummy:Dummy2007_2008 + FranceDummy:Dummy2009Onwards + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2009Onwards + FinlandDummy:Dummy2007_2008 + FinlandDummy:Dummy2009Onwards + gearingLag:CoreDummy + I(gearingLag^2):CoreDummy + BrentCrude + RealGDPGrowth + CPIGrowth:InflationSurprise + RealGDPGrowth:InflationSurprise +gearingLag:UpmarketDummy + gearingLag:DownMarketDummy , data=panel, model="pooling", index = c("FundID", "Year"))

＃between model“ between”对个人或时间均值进行估算

betweenmodel <- plm(TotalReturnAnnual~ UKDummy+ GermanyDummy +  NetherlandsDummy + FranceDummy + ItalyDummy + FinlandDummy + Residential + IndustrialLogistics + Multi.Sector + Office + Retail +  CoreClosedEnd+ GAV +  I(GAV^2)+ Age + I(Age^2)+ gearingLag  + I(gearingLag^2)+ InterestRate +CPIGrowth + BBBBonds + InflationSurprise + M1MoneySupply  + InterestRate + Dummy2009Onwards+ GermanyDummy:Dummy2007_2008 + UKDummy:Dummy2009Onwards + UKDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2009Onwards + GermanyDummy:Dummy2007_2008 + GermanyDummy:Dummy2009Onwards + FranceDummy:Dummy2007_2008 + FranceDummy:Dummy2009Onwards + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2009Onwards + FinlandDummy:Dummy2007_2008 + FinlandDummy:Dummy2009Onwards + gearingLag:CoreDummy + I(gearingLag^2):CoreDummy + BrentCrude + RealGDPGrowth + CPIGrowth:InflationSurprise + RealGDPGrowth:InflationSurprise +gearingLag:UpmarketDummy + gearingLag:DownMarketDummy, data=panel, model="between", index = c("FundID", "Year"))

随机模型

Europerandom <- plm(TotalReturnAnnual~ UKDummy+ GermanyDummy +  NetherlandsDummy + FranceDummy + ItalyDummy + FinlandDummy + Residential + IndustrialLogistics + Multi.Sector + Office + Retail +  CoreClosedEnd+ GAV +  I(GAV^2)+ Age + I(Age^2)+ gearingLag  + I(gearingLag^2)+ InterestRate +CPIGrowth + BBBBonds + InflationSurprise + M1MoneySupply  + InterestRate + Dummy2009Onwards+ GermanyDummy:Dummy2007_2008 + UKDummy:Dummy2009Onwards + UKDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2007_2008 + NetherlandsDummy:Dummy2009Onwards + GermanyDummy:Dummy2007_2008 + GermanyDummy:Dummy2009Onwards + FranceDummy:Dummy2007_2008 + FranceDummy:Dummy2009Onwards + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2007_2008 + ItalyDummy:Dummy2009Onwards + FinlandDummy:Dummy2007_2008 + FinlandDummy:Dummy2009Onwards + gearingLag:CoreDummy + I(gearingLag^2):CoreDummy + BrentCrude + RealGDPGrowth + CPIGrowth:InflationSurprise + RealGDPGrowth:InflationSurprise +gearingLag:UpmarketDummy + gearingLag:DownMarketDummy , data=panel, model="random", index = c("FundID", "Year"))

我希望能够自信地证明使用正确方法的正确性，对缺乏数据表示歉意，我无法共享它。