R語(yǔ)言常用上機(jī)命令分功能整理——時(shí)間序列分析為主

1、R語(yǔ)言常用上機(jī)命令分功能整理時(shí)間序列分析為主第一講應(yīng)用實(shí)例 R的基本界面是一個(gè)交互式命令窗口，命令提示符是一個(gè)大于號(hào)，命令的結(jié)果馬上顯示在命令下面。 S命令主要有兩種形式：表達(dá)式或賦值運(yùn)算（用<或者=表示）。在命令提示符后鍵入一個(gè)表達(dá)式表示計(jì)算此表達(dá)式并顯示結(jié)果。賦值運(yùn)算把賦值號(hào)右邊的值計(jì)算出來(lái)賦給左邊的變量。 可以用向上光標(biāo)鍵來(lái)找回以前運(yùn)行的命令再次運(yùn)行或修改后再運(yùn)行。 S是區(qū)分大小寫(xiě)的，所以x和X是不同的名字。我們用一些例子來(lái)看R軟件的特點(diǎn)。假設(shè)我們已經(jīng)進(jìn)入了R的交互式窗口。如果沒(méi)有打開(kāi)的圖形窗口，在R中，用：> x11() 可以打開(kāi)一個(gè)作圖窗口。然后，輸入以下語(yǔ)句： x1 =

2、 0:100 x2 = x1*2*pi/100 y = sin(x2) plot(x2,y,type="l") 這些語(yǔ)句可以繪制正弦曲線(xiàn)圖。其中，“=”是賦值運(yùn)算符。0:100表示一個(gè)從0到100 的等差數(shù)列向量。第二個(gè)語(yǔ)句可以看出，我們可以對(duì)向量直接進(jìn)行四則運(yùn)算，計(jì)算得到的x2 是向量x1的所有元素乘以常數(shù)2*pi/100的結(jié)果。從第三個(gè)語(yǔ)句可看到函數(shù)可以以向量為輸入，并可以輸出一個(gè)向量，結(jié)果向量y的每一個(gè)分量是自變量x2的每一個(gè)分量的正弦函數(shù)值。plot(x2,y, type="l",main="畫(huà)圖練習(xí)",sub="好

3、好練", xlab="x軸",ylab='y軸')有關(guān)作圖命令plot的詳細(xì)介紹可以在R中輸入help(plot)數(shù)學(xué)函數(shù)abs，sqrt：絕對(duì)值，平方根 log, log10, log2 , exp：對(duì)數(shù)與指數(shù)函數(shù) sin，cos，tan，asin，acos，atan，atan2：三角函數(shù) sinh，cosh，tanh，asinh，acosh，atanh：雙曲函數(shù) 簡(jiǎn)單統(tǒng)計(jì)量sum, mean, var, sd, min, max, range, median, IQR（四分位間距）等為統(tǒng)計(jì)量，sort，order，rank與排序有關(guān)，其它還有a

4、ve，fivenum，mad，quantile，stem等。下面我們看一看S的統(tǒng)計(jì)功能:> marks <- c(10, 6, 4, 7, 8) > mean(marks) > sd(marks) > min(marks)> max(marks) 第一個(gè)語(yǔ)句輸入若干數(shù)據(jù)到一個(gè)向量，函c()用來(lái)把數(shù)據(jù)組合為一個(gè)向量。后面用了幾個(gè)函數(shù)來(lái)計(jì)算數(shù)據(jù)的均值、標(biāo)準(zhǔn)差、最小值、最大值?？梢园讶舾尚忻畋４嬖谝粋€(gè)文本文件中，然后用source函數(shù)來(lái)運(yùn)行整個(gè)文件：> source("C:/l.R") 注意字符串中的反斜杠。例:計(jì)算6, 4, 7, 8

5、，10的均值和標(biāo)準(zhǔn)差，把若干行命令保存在一個(gè)文本文件（比如C:1.R）中，然后用source函數(shù)來(lái)運(yùn)行整個(gè)文件。a<- c(10, 6, 4, 7, 8) b<-mean(a) c<-sd(a) source("C:/1.R")時(shí)間序列數(shù)據(jù)的輸入使用函數(shù)tsts(1:10, frequency = 4, start = c(1959, 2) print( ts(1:10, frequency = 7, start = c(12, 2), calendar = TRUE)a<-ts(1:10, frequency = 4, start = c(1959

6、, 2)plot(a)將外部數(shù)據(jù)讀入Rread.csv默認(rèn)header = TRUE，也就是第一行是標(biāo)簽，不是數(shù)據(jù)。read.table默認(rèn)header = FALSE將R中的數(shù)據(jù)輸出write write.table write.csv第二講1. 繪制時(shí)序圖、自相關(guān)圖例題2.1 d=scan("sha.csv")sha=ts(d,start=1964,freq=1)plot.ts(sha) #繪制時(shí)序圖acf(sha,22) #繪制自相關(guān)圖，滯后期數(shù)22pacf(sha,22) #繪制偏自相關(guān)圖，滯后期數(shù)22corr=acf(sha,22) #保存相關(guān)系數(shù)cov=acf(

7、sha,22,type = "covariance") #保存協(xié)方差圖的保存，單擊選中圖，在菜單欄選中“文件”，再選“另存為”。同時(shí)顯示多個(gè)圖：用x11()命令生成一個(gè)空白圖，再輸入作圖命令。2. 同時(shí)繪制兩組數(shù)據(jù)的時(shí)序圖d=read.csv("double.csv",header=F)double=ts(d,start=1964,freq=1)plot(double, plot.type = "multiple") #兩組數(shù)據(jù)兩個(gè)圖plot(double, plot.type = "single") #兩組數(shù)據(jù)一

8、個(gè)圖 plot(double, plot.type = "single",col=c("red","green"),lty=c(1,2) #設(shè)置每組數(shù)據(jù)圖的顏色、曲線(xiàn)類(lèi)型)3.產(chǎn)生服從正態(tài)分布的隨機(jī)觀察值例題2.4 隨機(jī)產(chǎn)生1000白噪聲序列觀察值d=rnorm(1000,0,1) #個(gè)數(shù)1000 均值0 方差1plot.ts(d)4.純隨機(jī)性檢驗(yàn)例題2.3續(xù)d=scan("temp.csv")temp=ts(d,freq=1,start=c(1949)Box.test(temp, type="Ljung

9、-Box",lag=6)5.差分計(jì)算x=1:10y=diff(x)k步差分 加入?yún)?shù) lag=k如計(jì)算x的3步差分為y=diff(x, lag = 3)p階差分 加入?yún)?shù)differences = p如2階差分y=diff(x,differences = 2)第三講例題3.1plot.ts(arima.sim(n = 100, list(ar = 0.8)#模擬AR(1)模型，并作時(shí)序圖。plot.ts(arima.sim(n = 100, list(ar = -1.1)#非平穩(wěn)，無(wú)法得到時(shí)序圖。plot.ts(arima.sim(n = 100, list(ar = c(1,-0.

10、5)plot.ts(arima.sim(n = 100, list(ar = c(1,0.5)例題3.5acf(arima.sim(n = 100, list(ar = 0.8)acf (arima.sim(n = 100, list(ar = -1.1)acf (arima.sim(n = 100, list(ar = c(1,-0.5)acf (arima.sim(n = 100, list(ar = c(1,0.5)例題3.7arima.sim(n = 1000, list(ar = 0.5, ma = -0.8)acf(arima.sim(n = 1000, list(ar = 0.5

11、, ma = -0.8),20)pacf(arima.sim(n = 1000, list(ar = 0.5, ma = -0.8),20)例題2.5d=scan("a1.5.txt") #導(dǎo)入數(shù)據(jù)prop=ts(d,start=1950,freq=1) #轉(zhuǎn)化為時(shí)間序列數(shù)據(jù)plot(prop) #作時(shí)序圖acf(prop,12) #作自相關(guān)圖，拖尾pacf(prop,12) #作偏自相關(guān)圖，1階截尾Box.test(prop, type="Ljung-Box",lag=6) #純隨機(jī)性檢驗(yàn),p值小于5%，序列為非白噪聲Box.test(prop, ty

12、pe="Ljung-Box",lag=12)arima(prop, order = c(1,0,0),method="ML") #用AR(1)模型擬合，如參數(shù)method="CSS"，估計(jì)方法為條件最小二乘法，用條件最小二乘法時(shí)，不顯示AIC。arima(prop, order = c(1,0,0),method="ML", include.mean = F) #用AR(1)模型擬合，不含截距項(xiàng)。tsdiag(arima(prop, order = c(1,0,0),method="ML") #

13、對(duì)估計(jì)進(jìn)行診斷，判斷殘差是否為白噪聲summary(arima(prop, order = c(1,0,0),method="ML")a=arima(prop, order = c(1,0,0),method="ML")r=a$residuals#用r來(lái)保存殘差Box.test(r,type="Ljung-Box",lag=6)#對(duì)殘差進(jìn)行純隨機(jī)性檢驗(yàn)predict(arima(prop, order = c(1,0,0), n.ahead =5) #預(yù)測(cè)未來(lái)5期prop.fore = predict(arima(prop, orde

14、r = c(1,0,0), n.ahead =5)#將未來(lái)5期預(yù)測(cè)值保存在prop.fore變量中U = prop.fore$pred + 1.96* prop.fore$seL = prop.fore$pred 1.96* prop.fore$se#算出95%置信區(qū)間ts.plot(prop, prop.fore$pred,col=1:2)#作時(shí)序圖，含預(yù)測(cè)。lines(U, col="blue", lty="dashed")lines(L, col="blue", lty="dashed")#在時(shí)序圖中作出95

15、%置信區(qū)間例題3.9d=scan("a1.22.txt")x=diff(d)arima(x, order = c(1,0,1),method="CSS")tsdiag(arima(x, order = c(1,0,1),method="CSS")第一點(diǎn)：對(duì)于第三講中的例2.5，運(yùn)行命令arima(prop, order = c(1,0,0),method="ML")之后，顯示：Call:arima(x = prop, order = c(1, 0, 0), method = "ML")Coeff

16、icients: ar1 intercept 0.6914 81.5509s.e. 0.0989 1.7453sigma2 estimated as 15.51: log likelihood = -137.02, aic = 280.05注意：intercept下面的81.5509是均值，而不是截距！雖然intercept是截距的意思，這里如果用mean會(huì)更好。（the mean and the intercept are the same only when there is no AR term，均值和截距是相同的，只有在沒(méi)有AR項(xiàng)的時(shí)候）如果想得到截距，利用公式計(jì)算。int=(1-0.

17、6914)*81.5509= 25.16661。課本P81的例2.5續(xù)中的截距25.17是正確的。第二點(diǎn)：如需計(jì)算參數(shù)的t統(tǒng)計(jì)量值和p值，利用下面的公式。ar的t統(tǒng)計(jì)量值=0.6914/0.0989= 6.9909(注：數(shù)值與課本略有不同，因?yàn)檎n本用sas算的se= 0.1029，R計(jì)算的se=0.0989)p值=pt(6.9909,df=48,lower.tail = F)*2pt()為求t分布求p值的函數(shù)，6.99為t統(tǒng)計(jì)量的絕對(duì)值，df為自由度=數(shù)據(jù)個(gè)數(shù)-參數(shù)個(gè)數(shù)，lower.tail = F表示所求p值為PT > t，如不加入這個(gè)參數(shù)表示所求p值為PT <=t。乘2表示p值

18、是雙側(cè)的（課本上的p值由sas算出，是雙側(cè)的）均值的t統(tǒng)計(jì)量值和p值同理。在時(shí)間序列中對(duì)參數(shù)顯著性的要求與回歸模型不同，我們更多的是考察模型整體的好壞，而不是參數(shù)。所以，R中的arima擬合結(jié)果中沒(méi)有給出參數(shù)的t統(tǒng)計(jì)量值和p值，如果題目沒(méi)有特別要求，一般不需要手動(dòng)計(jì)算。第三點(diǎn)：修正第三講中的錯(cuò)誤：例2.5中，我們用下面的語(yǔ)句對(duì)擬合arima模型之后的殘差進(jìn)行了LB檢驗(yàn)：a=arima(prop, order = c(1,0,0),method="ML")r=a$residualsa=arima(prop, order = c(1,0,0),method="ML&q

19、uot;)r=a$residuals#用r來(lái)保存殘差Box.test(r,type="Ljung-Box",lag=6)#對(duì)殘差進(jìn)行純隨機(jī)性檢驗(yàn)最后一句不完整，需要加上參數(shù)fitdf=1，修改為Box.test(r,type="Ljung-Box",lag=6，fitdf=1)fitdf表示p+q，number of degrees of freedom to be subtracted if x is a series of residuals，當(dāng)檢驗(yàn)的序列是殘差到時(shí)候，需要加上命令fitdf，表示減去的自由度。運(yùn)行Box.test(r,type=&q

20、uot;Ljung-Box",lag=6,fitdf=1)后，顯示的結(jié)果：Box.test(r,type="Ljung-Box",lag=6,fitdf=1) Box-Ljung testdata: r X-squared = 5.8661, df = 5, p-value = 0.3195“df = 5”表示自由度為5，由于參數(shù)lag=6，所以是滯后6期的檢驗(yàn)。第四講# example4_1 擬合線(xiàn)性模型x1=c(12.79,14.02,12.92,18.27,21.22,18.81,25.73,26.27,26.75,28.73,31.71,33.95)a=a

21、s.ts(x1)is.ts(a)ts.plot(a)t=1:12tlm1=lm(at)summary(lm1) # 返回?cái)M合參數(shù)的統(tǒng)計(jì)量coef(lm1) #返回被估計(jì)的系數(shù)fitted(lm1) #返回模擬值residuals(lm1) #返回殘差值fit1=as.ts(fitted(lm1)ts.plot(a);lines(fit1,col="red") #擬合圖 #eg1cs=ts(scan("eg1.txt",sep=",")csts.plot(cs)t=1:40lm2=lm(cst)summary(lm2) # 返回?cái)M合參數(shù)

22、的統(tǒng)計(jì)量coef(lm2) #返回被估計(jì)的系數(shù)fit2=as.ts(fitted(lm2) #返回模擬值residuals(lm2) #返回殘差值ts.plot(cs);lines(fit2,col="red") #擬合圖 #example4_2 擬合非線(xiàn)性模型t=1:14x2=c(1.85,7.48,14.29,23.02,37.42,74.27,140.72,265.81,528.23,1040.27,2064.25,4113.73,8212.21,16405.95)x2plot(t,x2)m1=nls(x2a*t+bt,start=list(a=0.1,b=1.1),

23、trace=T)summary(m1) # 返回?cái)M合參數(shù)的統(tǒng)計(jì)量coef(m1) #返回被估計(jì)的系數(shù)fitted(m1) #返回模擬值residuals(m1) #返回殘差值plot(t,x2);lines(t,fitted(m1) #擬合圖#讀取excel中讀取文件，逗號(hào)分隔符 a=read.csv("example4_2.csv",header=TRUE)t=a$tx=a$xxts.plot(x)m2=nls(xa*t+bt,start=list(a=0.1,b=1.1),trace=T)summary(m2) # 返回?cái)M合參數(shù)的統(tǒng)計(jì)量coef(m2) #返回被估計(jì)的系

24、數(shù)fitted(m2) #返回模擬值 residuals(m2) #返回殘差值plot(t,x);lines(t,fitted(m2) #擬合圖#eg2I<-scan("eg2.txt")Ix=ts(data=I,start=c(1991,1),f=12) #化為時(shí)間序列 xplot.ts(x)t=1:130t2=t2m3=lm(xt+t2)coef(m3) #返回被估計(jì)的系數(shù) summary(m3) # 返回?cái)M合參數(shù)的統(tǒng)計(jì)量#去不顯著的自變量 ，再次模擬 m4=lm(xt2) coef(m4) #返回被估計(jì)的系數(shù)summary(m4) # 返回?cái)M合參數(shù)的統(tǒng)計(jì)量m2=

25、fitted(m4) #返回模擬值y=ts(data=m2,start=c(1991,1),f=12)yts.plot(x);lines(y)#平滑法 #簡(jiǎn)單移動(dòng)平均法x=c(5,5.4,5.8,6.2)xy=filter(x,rep(1/4,4),sides=1) y#指數(shù)平滑for(i in 1:3) x1=x1 xi+1=0.25*xi+1+0.75*xi #HoltWinters Filtera=ts(read.csv("holt.csv",header=F),start=c(1978,1),f=1)am=HoltWinters(a,alpha=0.15,beta=

26、0.1,gamma=FALSE,l.start=51259,b.start=4325)mfitted(m)plot(m)plot(fitted(m) #綜合cs=ts(read.csv("eg3.csv",header=F),start=c(1993,1),f=12) #讀取數(shù)據(jù) csts.plot(cs) #繪制時(shí)序圖 cs.sea1=rep(0,12)cs.sea1for(i in 1:12) for(j in 1:8) cs.sea1i=cs.sea1i+csi+12*(j-1) cs.sea=(cs.sea1/8)/(mean(cs)cs.seacs.sea2=re

27、p(cs.sea,8)cs.sea2x=cs/cs.sea2xplot(x)t=1:96m1=lm(xt)coef(m1)summary(m1) m=ts(fitted(m1),start=c(1993,1),f=12)ts.plot(x,type="p");lines(m,col="red")r=residuals(m1)Box.test(r) #白噪聲檢驗(yàn)第五講#回顧#例5.1sha=ts(scan("sha.csv"),start=1964,freq=1)ts.plot(sha)diff(sha)par(mfrow=c(2,1)

28、ts.plot(diff(sha)acf(diff(sha)#例5.2car=ts(read.csv("car.csv",header=F),start=1950,freq=1)carpar(mfrow=c(3,1)ts.plot(car)ts.plot(diff(car)ts.plot(diff(car,differences=2)#例5.3milk=ts(scan("milk.txt"),start=c(1962,1),freq=12)milkpar(mfrow=c(3,1)ts.plot(milk)ts.plot(diff(milk)dm1=dif

29、f(diff(milk),lag=12)ts.plot(dm1)acf(dm1)#例5.5x=ts(cumsum(rnorm(1000,0,100)ts.plot(x)#擬合ARIMA模型#5.8.1a=ts(scan("581.txt")par(mfrow=c(2,2)ts.plot(a)da=diff(a)ts.plot(da)acf(da,20)pacf(da,20)Box.test(da,6)fit1=arima(a,c(1,1,0),method="ML")predict(fit1,5)#incom=ts(read.csv("inco

30、m.csv",header=F),start=1952,freq=1)incomts.plot(incom)dincom=diff(incom)ts.plot(dincom)acf(dincom,lag=18) #自相關(guān)圖Box.test(dincom,type="Ljung-Box",lag=6) #白噪聲檢驗(yàn)Box.test(dincom,type="Ljung-Box",lag=12)Box.test(dincom,type="Ljung-Box",lag=18)pacf(dincom,lag=18)fit1=arim

31、a(dincom,order=c(0,0,1),method="CSS")fit2=arima(incom,order=c(0,1,1),xreg=1:length(incom),method="CSS") #見(jiàn)/stoffer/tsa2/Rissues.htmBox.test(fit2$resid,lag=6,type="Ljung-Box",fitdf=1)fore=predict(fit2,10,newxreg=(length(incom)+1):(length(incom)+10)

32、#疏系數(shù)模型#例5.8w=ts(read.csv("w.csv"),start=1917,freq=1)w=w,1par(mfrow=c(2,2)ts.plot(w)ts.plot(diff(w)acf(diff(w),lag=18)pacf(diff(w),lag=18)dw=diff(w)fit3=arima(dw,order=c(4,0,0),fixed=c(NA,0,0,NA,0),method="CSS")Box.test(fit3$resid,lag=6,type="Ljung-Box",fitdf=2)Box.test(

33、fit3$resid,lag=12,type="Ljung-Box",fitdf=2)fit4=armaFit(arima(4,0,0),fixed=c(NA,0,0,NA),include.mean=F,data=dw,method="CSS")summary(fit4)#例 5.9ue=ts(scan("unemployment.txt"),start=1962,f=4) #讀取數(shù)據(jù)par(mfrow=c(2,2) #繪制時(shí)序圖ts.plot(ue)#差分due=diff(ue)ddue=diff(due,lag=4)ts.plo

34、t(ddue)Box.test(ddue,lag=6)#平穩(wěn)性檢驗(yàn)acf(ddue,lag=30)pacf(ddue,lag=30)arima(ddue,order=c(0,0,0),method="ML")arima(ddue,order=c(4,0,0),method="ML")arma=arima(ddue,order=c(4,0,0),transform.pars=F,fixed=c(NA,0,0,NA),include.mean=F,method="ML")#參數(shù)估計(jì)與檢驗(yàn)（加載fArma程序包）fit2=armaFit(a

35、rima(4,0,0),include.mean=F,data=ddue,method="ML")summary(fit2)fit3=armaFit(arima(4,0,0),data=ddue,transform.pars=F,fixed=c(NA,0,0,NA),include.mean=F,method="CSS")summary(fit3)#殘差白噪聲檢驗(yàn)Box.test(arma$resid,6,fitdf=2,type="Ljung")#擬合ft=ts(fitted(fit3),start=1963.25,f=4)dft=

36、ts(rep(0,115),start=1963.25,f=4)for(i in 1:115)dfti=fti+duei+uei+4ts.plot(ue);lines(dft,col="red") #例5.10 乘積季節(jié)模型wue=ts(scan("wue.txt"),start=1948,f=12)arima(wue,order=c(1,1,1),seasonal=list(order=c(0,1,1),period=12),include.mean=F,method="CSS")#擬合Auto-Regressive模型eg1=ts

37、(scan("582.txt")ts.plot(eg1)#因變量關(guān)于時(shí)間的回歸模型fit.gls=gls(eg1-1+time(eg1),correlation=corARMA(p=1),method="ML")#see the nlme packagesummary(fit.gls2) #the results#延遲因變量回歸模型leg1=lag(eg1,-1)y=cbind(eg1,leg1)fit=arima(y,1,c(0,0,0),xreg=y,2,include.mean=F)第六講#回顧#例5.1sha=ts(scan("sha.

38、csv"),start=1964,freq=1)ts.plot(sha)diff(sha)par(mfrow=c(2,1)ts.plot(diff(sha)acf(diff(sha)#例5.2car=ts(read.csv("car.csv",header=F),start=1950,freq=1)carpar(mfrow=c(3,1)ts.plot(car)ts.plot(diff(car)ts.plot(diff(car,differences=2)#例5.3milk=ts(scan("milk.txt"),start=c(1962,1),

39、freq=12)milkpar(mfrow=c(3,1)ts.plot(milk)ts.plot(diff(milk)dm1=diff(diff(milk),lag=12)ts.plot(dm1)acf(dm1)#例5.5x=ts(cumsum(rnorm(1000,0,100)ts.plot(x)#擬合ARIMA模型#上機(jī)指導(dǎo)5.8.1a=ts(scan("581.txt")par(mfrow=c(2,2)ts.plot(a)da=diff(a)ts.plot(da)acf(da,20)pacf(da,20)Box.test(da,6)fit1=arima(a,c(1,1

40、,0),method="ML")predict(fit1,5,newxreg=(length(a)+1):(length(a)+5)fit2=armaFit(arima(1,1,0),data=a,xreg=1:length(a),method="ML")summary(fit1)summary(fit2)#截距項(xiàng)不顯著，故舍去fit3=arima(a,c(1,1,0),method="ML") predict(fit3,5)#例5.8 incom=ts(read.csv("incom.csv",header=F)

41、,start=1952,freq=1)incomts.plot(incom)dincom=diff(incom)ts.plot(dincom)acf(dincom,lag=18) #自相關(guān)圖Box.test(dincom,type="Ljung-Box",lag=6) #白噪聲檢驗(yàn)pacf(dincom,lag=18)fit=arima(incom,order=c(0,1,1),xreg=1:length(incom),method="CSS") #見(jiàn)/stoffer/tsa2/Rissues.htmAuto

42、corTest(fit$resid) #加載FinTS包 fore=predict(fit,10,newxreg=(length(incom)+1):(length(incom)+10)#疏系數(shù)模型#例5.8w=ts(read.csv("w.csv"),start=1917,freq=1)w=w,1par(mfrow=c(2,2)ts.plot(w)ts.plot(diff(w)acf(diff(w),lag=18)pacf(diff(w),lag=18)dw=diff(w)fit3=arima(dw,order=c(4,0,0),fixed=c(NA,0,0,NA,0),

43、method="CSS")Box.test(fit3$resid,lag=6,type="Ljung-Box",fitdf=2)Box.test(fit3$resid,lag=12,type="Ljung-Box",fitdf=2)fit4=armaFit(arima(4,0,0),fixed=c(NA,0,0,NA),include.mean=F,data=dw,method="CSS") #加載fArma包 ，檢驗(yàn)參數(shù) summary(fit4)#例 5.9#讀取數(shù)據(jù)ue=ts(scan("unemp

44、loyment.txt"),start=1962,f=4)#繪制時(shí)序圖par(mfrow=c(2,2)ts.plot(ue)#差分due=diff(ue)ddue=diff(due,lag=4)ts.plot(ddue)Box.test(ddue,lag=6)#平穩(wěn)性檢驗(yàn)acf(ddue,lag=30)pacf(ddue,lag=30)arima(ddue,order=c(0,0,0),method="ML")arima(ddue,order=c(4,0,0),method="ML")arma=arima(ddue,order=c(4,0,0),transform.pars=F,fixed=c(NA,0,0,NA),include.mean=F,method="ML")#參數(shù)估計(jì)與檢驗(yàn)（加載fArma程序包）fit2=armaFit(arima(4,0,0),include.mean=F,data=ddue,method="ML")summary(fit2)fit3=armaFit(arima(4,0,0),data=ddue,transform.pars=F,fixed=c(NA,0,0,NA),include.mean=F,method=&