應(yīng)用時(shí)間序列分析實(shí)訓(xùn)報(bào)告

1、應(yīng)用時(shí)間序列分析實(shí)訓(xùn)報(bào)告實(shí)訓(xùn)項(xiàng)目名稱非平穩(wěn)時(shí)間序列模型的建立實(shí)訓(xùn)時(shí)間 2013年12月16日實(shí) 訓(xùn)地點(diǎn) 實(shí)驗(yàn)樓308班 級(jí)計(jì)科1001班學(xué)號(hào)姓名應(yīng)用時(shí)間序列分析實(shí)訓(xùn)（實(shí)踐）報(bào)告實(shí)訓(xùn)名稱平穩(wěn)時(shí)間序列模型的建立一、實(shí)訓(xùn)目的本次實(shí)驗(yàn)是一個(gè)綜合試驗(yàn)，通過自己選定問題，收集數(shù)據(jù)， 確定研究方法，建立合適模型，解決實(shí)際問題，增強(qiáng)學(xué)生 動(dòng)手能力，提高學(xué)生綜合分析的能力。二、實(shí)訓(xùn)內(nèi)容學(xué)生根據(jù)自己喜好，選定一個(gè)實(shí)際問題，確定指標(biāo)，收集相關(guān)數(shù)據(jù)，利用所 學(xué)時(shí)間序列分析方法隊(duì)進(jìn)行研究，建立時(shí)間序列模型，揭示其研究對(duì)象內(nèi)部的規(guī) 律，并對(duì)未來進(jìn)行預(yù)測(cè)。并寫出分析報(bào)告。具體實(shí)驗(yàn)內(nèi)容如下：1確定研究問題2收集數(shù)據(jù)3建立合適

2、模型ARIMA模型建模前的準(zhǔn)備：判斷序列是否平穩(wěn).通過序列自相關(guān)圖、趨勢(shì)圖等進(jìn)行判斷若序列不平穩(wěn)：均值非平穩(wěn)序列通過差分變換轉(zhuǎn)換為平穩(wěn)方差非平穩(wěn)序列通過對(duì)數(shù)變換等轉(zhuǎn)化為平穩(wěn)序列模型平穩(wěn)化以后，將序列零均值化模型識(shí)別主要通過序列的自相關(guān)函數(shù)、偏自相關(guān)函數(shù)表現(xiàn)的特征，進(jìn)行初步的模型識(shí) 別3 .模型參數(shù)估計(jì)在Eviews中估計(jì)ARMA模型的方法估計(jì)模型以后要能寫出模型的形式（差分方程形式和用B算子表示的形 式）模型的診斷檢驗(yàn)根據(jù)模型殘差是不是白噪聲來判斷模型是否為適應(yīng)性模型能根據(jù)輸出結(jié)果判斷模型是否平穩(wěn)，是否可逆若有多個(gè)序列是模型的適應(yīng)性模型，會(huì)用合適的方法從這些模型中進(jìn)行選 擇，如比較模型的殘差方

3、差，AIC，SC等。模型應(yīng)用掌握追溯預(yù)測(cè)的操作方法外推預(yù)測(cè)的操作方法四、實(shí)訓(xùn)分析與總結(jié)1）輸入數(shù)據(jù)2）生成時(shí)序圖觀測(cè)序歹觀測(cè)差分時(shí)序圖看出并無明顯的趨勢(shì)性或者循環(huán)性，得出一階差分平穩(wěn)。1955 1960 1965 1970 1975 1900 1905D(X) 33 File Edi t. Object. Vi ew Proc Quick Option.! Window HelpA&m I PrcicgljjE匚11 Pr叩航國(guó) 更int |N曰me | Freeze| 5曰mple | Genr |SIieeI:1|Ident |Line| Emr |Correlogram of D(X)D

4、ate /16/13 Time: 11:39Sarr. 1952 1988Included obsen/ations: 36AutocorrelationPartial CorrelationACPAC Q-Stat ProbZl ii匚i i.Zl i序列階自相關(guān)顯著,123456891011121314J序0.5350.2020.021-0.202-0.192-0.108-0.0170.1360.1490.037-0.0290.020-0.010-0.0560.535-0.117-0.052-0.2400.0570.0030.0550.111-0.025-0.104-0.0180.164-

5、0.042-0.079;Q統(tǒng)計(jì)11.13112.82812.84714.58416.20316.73716.75017.65818.78918.8600.0010.0020.0050.0060.0060.0100.0190.0240.0270.04218.908 0.06318.93118.93719.1320.0900.1250.160值小于0.05,非白噪聲;由圖知,.同時(shí)偏 自相關(guān)拖尾、自相關(guān) 步截尾，可建立 ARIMA（0,1,1）模型。3）模型參數(shù)估計(jì)File Edi t Object Vi ew Proc Quick Otiore Window HelpViEVj|rciiz|c）

6、bjEizt| 叫11亦血舊能況| EstimmE|口佗匚禎| 沮出唱也|Dependent Variable: D(X)Method: Least SquaresDate: 12/16/13 Time: 11:43Sample (adjusted): 1953 1988Included observations: 36 after adjustmentsConvergence achieved after 11 iterationsBackcast: 1952VariableCoefficientStd. Error t-StatisticProb.C5.0155662.1302702.35

7、44280.0245MA(1)7081690.1263675.604044.R-squared0.318165Mean dependent var4.983333Adjusted R-squared0.298111S.D. dependent var8.970762S.E. of regression7.515597Akaike info criterion6.925791Sum squared resid1920.463Schwarz criterion7.013764Log likelihood-122.6642F-statistic15.86545ARMA模型估計(jì)方懇：MDProb(F-

8、statistic)0.000340M 舔曲 5.015566 甘(1 + 0.708169B )e t=tSBC 值為 7.013764m? t*ieTSquation: UNTITLED Torkfile:例5.3 File Edit Object View Proc Quick Otions Window Help訕已|穴匚|口可已匚1:| Print | Name | Freeze Estimate Forecast | Stats |Resids|Dependent Variable: D(X)Method: Least SquaresDate: 12/16/13 Time: 11:

9、45Sample (adjusted): 1954 1988Included observations: 35 after adjustmentsConvergence achieved after 3 iterationsVariableCoefficientStd. Error t-StatisticProb.C5.2331202.8430581.8406660.0747AR(1)0.5381160.1463483.6769520.0000R-squared0.290627Mean dependent var5.080000Adjusted R-squared0.269131S.D. de

10、pendent var9.082606S.E. of regression7.764066Akaike info criterion6.992541Sum squared resid1989.674Schwarz criterion7.081410Log likelihood-120.3695F-statistic13.51990Durbin-Watson stat1.879374Prob(F-statistic)0.000833Inverted AR Roots .54由圖知偏自相關(guān)，C的值大于0.05,則去掉C,繼續(xù)建立模型:ICTSquation:orJLeI File Edi t Ob

11、j ect Vi ew Proc Quick Otions WirLilow HelpView j Proc | Object | Print| Name I Freeze Estimate | Forecast! 5tats I ResidslDependent Variable: D(K)Method: Least SquaresDate: 12/16/13 Time: 11:47Sample (adjusted): 1954 1988Included observations: 35 after adjustmentsARIMA模型估計(jì)方程： = 0652119B Convergence

12、 achieved after 2 iterationsVariableCoefficientStd. Error t-StatisticProb.AR(1)0.6521190.1311164.9735930.0000R-squared0.234740Mean dependent var5.080000Adjusted R-squared0.234740S.D. dependent var9.082686S.E. of regression7.945457Akaike info criterion7.011233Sum squared resid2146.430Schwarz criterio

13、n7.055671Log likelihood-121.6966Durbin-Watson stat1.944132Inverted AR Roots.651SBC 值為 7.055671比較兩個(gè)模型的SBC值，建立ARMA模型最優(yōu)。4）模型的診斷檢驗(yàn)& EViews - Equation: UNTITLED Torkfile:例5-6國(guó)民收入指熟)UntitledI 】File Edi t Object View Proc Quick OE.ti ons Window HelpView|Proc|Object| 叩11亦邨舊能況| Estimate|Forecast|5t日愴|昭5成|Cor

14、relogram of ResidualsDate: 12/16/13 Time: 11:52Sample: 1954 1988Included observations: 35Q-statistic probabilities adjusted for 1 ARMA term(s)Autocorrelation Partial Correlation AC PAC Q-Stat ProbII I11 -0.025 -0.025 0.0232匚II 112 -0.111 -0.112 0.5096I113 0.056 0.051 0.6365匚I1匚14 -0.185 -0.198 2.069

15、9匚I1 115 -0.112 -0.112 2.6077I1匚16 -0.066 -0.129 2.0038匚I1匚17 -0.134 -0.165 3.6364 I118 0.133 0.070 4.4901 I11 19 0.159 0.100 5.7519I1110 -0.019 -0.018 5.7707I1匚111 -0.070 -0.135 6.0380I1112 0.079 0.057 6.3862I1113 -0.013 0.019 6.39650.4750.7270.5580.6250.7300.7260.7220.6750.7630.8120.8460.095殘差分析：P

16、值都大于01.05，顯著有效，是白噪聲序列。ID |15 -0.001 0.002 7.0378 0.933|16 -0.010 -0.007 7.0445 0.956也4U.bbbbUy19557.0913971956-0.34737919570.3394031958-2.1476301959-19.030421960-4.153247196112.1860319623.28266919637.59152219646.34838219652.52244719661.8266861967-3.5690751968-4.09987519692.29551119707.2087281971-2.7561081972-2.934663197313.578181974-2.1167051975-U.EI779291976-5.2215701977-1.91321819788.70847819796.3220711980-9