施三支,閆 麗,王澤升
(1.長(zhǎng)春理工大學(xué) 理學(xué)院,長(zhǎng)春 130022;2.吉林大學(xué) 數(shù)學(xué)學(xué)院,長(zhǎng)春 130012)
時(shí)間序列模型,特別是隨時(shí)間變化參數(shù)的向量自回歸(TVP-VAR)模型,在經(jīng)濟(jì)、金融等領(lǐng)域應(yīng)用廣泛[1-4].對(duì)隨時(shí)間變化參數(shù)的自回歸模型(TVPAR模型),目前采用的普遍方法是將所有的參數(shù)(包括隨機(jī)誤差)都視為隨機(jī)游動(dòng)過(guò)程.由于隨機(jī)游動(dòng)過(guò)程是非平穩(wěn)的,因此,當(dāng)隨機(jī)誤差被視為隨機(jī)游動(dòng)過(guò)程在分析金融數(shù)據(jù)的長(zhǎng)期變化時(shí)是不可靠的.文獻(xiàn)[5-6]用Markov鏈Monte Carlo(MCMC)方法討論了TVP-VAR模型中隨機(jī)誤差不變的情形.本文主要考慮p維TVPAR模型,當(dāng)變系數(shù)仍為系數(shù)已知的時(shí)間序列,而方差為同方差,且與p維時(shí)間序列模型的誤差方差均為已知時(shí),采用極大似然估計(jì)方法,對(duì)p維TVPAR模型中的未知參數(shù)進(jìn)行估計(jì),給出了估計(jì)的顯式表達(dá)式,并進(jìn)行了模擬計(jì)算.
p維TVPAR模型具有如下形式:
(1)
這里{xt,t=1,2,…,n}為可觀察數(shù)據(jù).本文要求模型滿(mǎn)足下列條件:
2)βi,ui是系數(shù)參數(shù),且-1<βi<1(i=1,2,…,p)未知,-1 3)βt,i(i=1,2,…,p;t=1,2,…,n)為不可觀測(cè)數(shù)據(jù);βt,1,βt,2,…,βt,p不相關(guān); 4)β0,i=0(i=1,2,…,p);x0=0. 由模型(1)知: 記β=(β1,β2,…,βp)T,u=(u1,u2,…,up)T.由假設(shè)條件1)知 即有 (2) 根據(jù)式(2)知,似然函數(shù)為 (3) 用類(lèi)似的方法可以討論模型(1)的平穩(wěn)性:記Xt=(xt,xt-1,…,xt-p+1)T,Xt是p×1維隨機(jī)向量.記Bt=(βt,1,βt,2,…,βt,p),則Bt為1×p維AR(1)過(guò)程,Bt各元素之間相互獨(dú)立.記U=diag(u1,u2,…,up),ηt=(ηt,1,ηt,2,…,ηt,p),將模型(1)改寫(xiě)為如下形式: Xt=(M+Dt)Xt-1+εt,Dt=Dt-1U+μt, (4) 表1 λ=0.1時(shí)不同p值下參數(shù)β的真值與估計(jì)值的比較Table 1 True values and estimated values of β under different p when λ=0.1 表2 λ=0.01時(shí)不同p值下參數(shù)β的真值與估計(jì)值的比較Table 2 True values and estimated values of β under diffrent p when λ=0.01 [1] Primiceri G E.Time Varying Structural Vector Autoregressions and Monetary Policy [J].Review of Economic Studies,2005,72(3):821-852. [2] Benati L,Mumtaz H U S.Evolving Macroeconomic Dynamics:A Structural Investigation [R].Working Paper 746.Frankfurt:European Central Bank,2007. [3] Baumeister C,Durinck E,Peersman G.Liquidity,Inflation and Asset Prices in a Time-Varying Framework for the Euro Area [J].National Bank of Belgium in Its Series Research Series,2008:10-17. [4] Nakajima J,Kasuya M,Watanabe T.Bayesian Analysis of Time-Varying Parameter Vector Autoregressive Model for the Japanese Economy and Monetary Policy [J].Journal of the Japanese and International Economics,2011,25(3):225-245. [5] Canova F,Gambetti L,Pappa E.The Structural Dynamics of Output Growth and Inflation:Some International Evidence [J].The Economic Journal,2007,117(3):167-191. [6] Gambetti L,Pappa E,Canova F.The Structural Dynamics of US Output and Inflation:What Explains the Changes [J].Journal of Money,Credit,and Banking,2006,40(2/3):369-388. [7] Weiss A A.The Stability of the AR(1) Process with an AR(1) Coefficient [J].Journal of Time Series Analysis,1985,6(3):181-186. [8] SHI San-zhi,WANG De-hui,SONG Li-xin,et al.Bayesian Estimation of Time Vary Parameter Autoregression Model [J].Journal of Jilin University:Science Edition,2011,49(5):857-860.(施三支,王德輝,宋立新,等.隨時(shí)間變化系數(shù)參數(shù)AR模型的Bayes估計(jì) [J].吉林大學(xué)學(xué)報(bào):理學(xué)版,2011,49(5):857-860.) (責(zé)任編輯:趙立芹) 研究簡(jiǎn)報(bào)1 模型的極大似然估計(jì)
2 模擬計(jì)算