伍德里奇計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論之計(jì)算機(jī)操作題的R語言實(shí)現(xiàn)(簡單線性回歸)

1、(1)(1)伍德里奇計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論之計(jì)算機(jī)操作題的R語言實(shí)現(xiàn)(簡單線性回歸)引言本學(xué)期正在學(xué)習(xí)伍德里奇第四版的計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論，為了增強(qiáng)對計(jì)量經(jīng)濟(jì)學(xué)的理解和熟悉R語言的操作，決定對計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論的計(jì)算機(jī)操作部分進(jìn)行R語言實(shí)現(xiàn)，如果各位看官發(fā)現(xiàn)有任何錯(cuò)誤請?jiān)谙旅嬖u(píng)論，不勝感激。題目就不在博客中重復(fù)，內(nèi)容主要是R語言代碼和習(xí)題解答。這一系列博客主要參照了計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論習(xí)題解答、王斌會(huì)老師出的計(jì)量經(jīng)濟(jì)學(xué)模型及R語言應(yīng)用以及AppliedEconometricswithRo第四版計(jì)量經(jīng)濟(jì)學(xué)導(dǎo)論exceI文件見經(jīng)過整理后的本博客數(shù)據(jù)和本篇博客Rmd文本見C2.1(1)#導(dǎo)入數(shù)據(jù)請注意修改文件路徑名data

2、_2.1-read.csv(E:博客計(jì)量經(jīng)濟(jì)學(xué)ch_2401k.csv,header=TRUE)#平均參與率ave_prate-mean(data_2.1$prate)#平均匹配率ave_mrate-mean(data_2.1$mrate)(2)#估計(jì)回歸方程lm_prate-lm(pratemrate,data=data_2.1)summary(lm_prate)#估計(jì)結(jié)果#prate=83.0755+5.8611mrate#n=1534#R-squared=0.0747(3)#解釋系數(shù)#當(dāng)匹配率為0的時(shí)候，回報(bào)率為83.0755%#mrate的系數(shù)表明在其他條件不變的條件下，當(dāng)匹配率每增加

3、一美元，回報(bào)率平均增加5.86%。(4)#輸入mrate數(shù)據(jù)輸出fit的值就是點(diǎn)估計(jì)point-data.frame(mrate=3.5)predict(lm(pratemrate,data=data_2.1),point,interval=confidence)#點(diǎn)估計(jì)為103.5892%,顯然這個(gè)答案是不可能的，這是由于數(shù)據(jù)集1534個(gè)記錄中，只有34個(gè)記錄中的mrate是大于3.5的，當(dāng)mrate較大時(shí)，擬合效果不好。(5)#prate的變異中，只有7.47%由mrate進(jìn)行解釋。該值不大，說明有其他因素影響著回報(bào)率。2.2(1)(1)data_2.2-read.csv(E:博客計(jì)量經(jīng)濟(jì)

4、學(xué)ch_2ceosal2.csv)#為簡便，將數(shù)據(jù)導(dǎo)入內(nèi)存attach(data_2.2)#平均年薪ave_salary-mean(salary)#平均任期ave_ceotenv-mean(ceoten)(2)#任期一年個(gè)數(shù)n_ceoten-length(which(ceoten=0)#最長任期max_ceoten-max(ceoten)(3)#參數(shù)估計(jì)lm_ceosal-lm(log(salary)ceoten)summary(lm_ceosal)#回歸方程為log(salary)=6.505498+0.009723*ceoten#報(bào)告#每多擔(dān)任一年ceo,年薪平均增長幾乎1%#釋放數(shù)據(jù)集d

5、etach(data_2.2)C2.3(1)data_2.3-read.csv(E:博客計(jì)量經(jīng)濟(jì)學(xué)ch_2sleep75.csv)attach(data_2.3)#方程擬合lm_sleep75-lm(sleeptotwrk)summary(lm_sleep75)#sleep=3586.37695-0.15075*totwrk#觀測次數(shù)#n=706#R-squared=0.1033#截距項(xiàng)表示，每周不工作時(shí)睡眠時(shí)間約等于3586.37695分鐘(2)#預(yù)測多工作兩小時(shí)(120分鐘)睡眠時(shí)間減少#Asleep=-0.15075*120=-18.09每晚減少-18.09分鐘，效應(yīng)不大detach(d

6、ata_2.3)2.4(2)(5)data_2.4-read.csv(E:博客計(jì)量經(jīng)濟(jì)學(xué)ch_2wage2.csv)attach(data_2.4)#平均工資ave_wage-mean(wage)#平均IQave_IQ-mean(IQ)#IQ樣本標(biāo)準(zhǔn)差sd_IQ-sd(IQ)(2)lm_wage2-lm(wageIQ)summary(lm_wage2)#回歸方程為#wage=116.9916+8.3031*IQ#n=935,R-squared=0.09554#IQ增加15單位，工資預(yù)期變化#Awage=8.3031*15=124.5465#IQ能夠解釋9.554%的工資變化(3)#由題可得，可

7、以構(gòu)建對數(shù)水平模型#log(wage)=a+bIQ+u#估計(jì)可得lm_wage-lm(log(wage)IQ)summary(lm_wage)#log(wage)=5.8869943+0.0088072*IQ#IQ提高15個(gè)單位wage平均提高13.21%#4og(wage)=0.0088072*15=0.132108detach(data2.4)C2.5(1)data_2.5-read.csv(E:博客計(jì)量經(jīng)濟(jì)學(xué)ch_2rdchem.csv)attach(data_2.5)#構(gòu)建模型為#log(rd)=a+b*log(sales)+u#匕為rd對sales的彈性(2)lm_rdchem-lm

8、(log(rd)log(sales)summary(lm_rdchem)#log(rd)=-4.10472+1.07573*log(sales)#彈性值為1.07573#當(dāng)其他條件不變時(shí)，sales每增口1%,rd增口1.07573%。detach(data_2.5)C2.6(1)data_2.6-read.csv(E:博客計(jì)量經(jīng)濟(jì)學(xué)ch_2Wmeap93.csv)attach(data_2.6)#隨著更多一美元的花費(fèi)，對通過率的影響越來越小。#因?yàn)楫?dāng)支持較低時(shí)，可以通過購買教學(xué)書籍、上輔導(dǎo)班，增加一美元對數(shù)學(xué)通過率提高影響較大；隨著支出增加到一定地步時(shí)，此時(shí)：、有名師(2)(5)(2)(5)

9、#見書中證明，略(3)(4)(5)C2.7(1)(2)(3)(4)lm_meap93-lm(math10Tog(expend)summary(lm_meap93)#math10=-69.341+11.164*log(expend)#n=408R-squared=0.02727#支出影響多大#當(dāng)其他條件不變時(shí)，expend每增加1%,math10平均增加0.11164%#支出提高10%,math10會(huì)增加1.1164%。max_fit-69.341+11.164*max(log(expend)detach(data_2.6)#可知數(shù)據(jù)集中最大為30.15033%,沒有超過100%data_2.7

10、-read.csv(E:博客計(jì)量經(jīng)濟(jì)學(xué)ch_2charity.csv)attach(data_2.7)#平均捐款數(shù)量ave_gift-mean(gift)#沒有捐款的百分比占比60%percent_gift-100*(length(which(respond=3)/length(respond)#平均郵遞數(shù)ave_mailsyear-mean(mailsyear)#最小、最大郵遞數(shù)min_mailsyear-min(mailsyear)max_mailsyear-max(mailsyear)lm_gift-lm(giftmailsyear)summary(lm_gift)#估計(jì)的回歸方程為#gift=2.0141+2.6495*mailsyear#n=4268,R-squared=0.01379#回歸系數(shù)表明當(dāng)其他條件不變時(shí)，每年的郵件郵寄數(shù)量每增加一封，捐款平均增加2.6495荷蘭盾#當(dāng)郵遞成本為1荷蘭盾時(shí)，仍然有2.6495