看了以上各位的,个人感觉如下:
1) Panel Data方面近年似乎没出更经典的教材(Ie:所以没进一步的理论更新),Econometric Analysis of Panel Data Baltagi那本算比较前沿的,尽管方法颇多但(Nonstationary PanelData;Unbalanced PanelData;etc)似乎仍还处于理论估计+初步实证状态;外文文献包括Elsevier.etc所用的也多是比较传统的(One-Way,Two-way,Autoregression)方法
2) 看到上面的XD说Kalman Filtering和Neuro Network,这些应该算金工的方法,而且例如Kalman Filtering(包括EKF UKF)都发展得比较成熟,个人感觉Kalman Filtering不大算前沿方法,只能说国内还感觉新奇。
Neuro Network潜力倒是很大,除了老套的BP,大家更喜欢用新奇的SOFM Hopfield ART那些,个人感觉这门分支的发展更多依靠计算机或者程序的发展,而经济学所要做的只是借鉴这块。
3) 个人感觉相对更新的是非线性方法(比如求函数相关性的Copula(eliptical Copula, Archemedian Copula),Markov,配合微分方程的研究,非参数...)以及估计方法的更新---Robust Estimation等
1:moment inequality 以前的矩估计都是考虑等式成立,局限性很大,而事实问题复杂的多。你看看现在顶级经济杂志,如econometrica,journal of econometrics 和 统计的annals ,jasa等。
2:statistical learning or machine learning, variable selection 国外很多牛人在做,国内刚刚开始。
1) dynamic panel: panel cointegration, cross sectional dependence, bootstrap.....
2) nonlinear dynamics: time series, multivariate, nonlinear ECM(STAR, THR, MS models......), nonlinear transformation of I(1)
3) long memory I(d) process with 0<d<1, robustness of unit root and cointegration test( against nonlinearity), long run variance(kernal, bandwith selection)......
4) nonparametric test of cointegration, nonparametric nonlinear......... continuous stochastic models of second moments
5) local time of brownian motion and its estimation
there are some nontrivial problems for the theoretical foundation of nonlinearity...................
Peter and Park's work confront the nontrivial problem of identification, it requires knowledge of analysis.................
Factor models are vary hot issues..
copula, stochasitc correlation are the beloved of finance people


雷达卡



京公网安备 11010802022788号







