Stacking Regressions: Latex Tables with R and stargazer

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

In my paper on the impact of the shale oil and gas boom in the US, I run various instrumental variables specifications. For these, it is nice to stack the regression results one on the other – in particular, to have one row for the IV results, one row for the Reduced Form and maybe one row for plain OLS to see how the IV may turn coefficients around.

I found that as of now – there is no way to do that directly; please correct me if I am wrong.

The layout I have in mind is as in the screenshot of my table.

In Stata, this can be accomplished through the use of esttab with the fragment, and append feature. This appends successive rows to an existing table and you can label the rows using the “refcat” option.

However, in R this is not possible as of yet. I have mainly worked with stargazer, as Marek has added felm objects for high dimensional fixed effects to be handled by his stargazer package.

The following functions are a “hack” that extracts the particular rows from the generated latex code by the stargazer package.

You can then piece the table together by combining the individual elements. The idea is that you have a single stargazer command that is passed to the various functions that extract the different features. Obviuously, this can be done a lot more elegant as the code is extremely hacky, but it does work

Marek has said that he is thinking of incorporating a stacking option into stargazer, but for now, my hack works reasonably well. The key thing to realise is that stargazer calls have the option to return

The following character strings can be used in the table.layout and omit.table.layout arguments of the stargazer command.

“-”	single horizontal line
“=”	double horizontal line
“-!”	mandatory single horizontal line
“=!”	mandatory double horizontal line
“l”	dependent variable caption
“d”	dependent variable labels
“m”	model label
“c”	column labels
“#”	model numbers
“b”	object names
“t”	coefficient table
“o”	omitted coefficient indicators
“a”	additional lines
“n”	notes
“s”	model statistics

The following functions will simply extract the rows that are being returned.

1. stargazer.keepcolnums<-function(call="") {
   
2. command<-gsub("\\)$",", table.layout='#'\\)",command)
   
3. call<-eval(parse(text=command))
   
4. bounds<-grep("tabular",call)
   
5. row1<- call[(bounds[1]+1):(bounds[2]-1)]
   
6. row1<-gsub("^\\\\\\\\\\[-1\\.8ex\\]\\\\hline $", "", row1)
   
7. row1<-gsub("^\\\\hline \\\\\\\\\\[-1\\.8ex\\] $","",row1)
   
8. row1<-row1[row1!=""]
   
9. }
   
10. stargazer.keeprow<-function(call="") {
    
11. command<-gsub("\\)$",", table.layout='t'\\)",command)
    
12. call<-eval(parse(text=command))
    
13. bounds<-grep("tabular",call)
    
14. row1<- call[(bounds[1]+1):(bounds[2]-1)]
    
15. row1<-gsub("^\\\\\\\\\\[-1\\.8ex\\]\\\\hline $", "", row1)
    
16. row1<-gsub("^\\\\hline \\\\\\\\\\[-1\\.8ex\\] $","",row1)
    
17. row1<-row1[row1!=""]
    
18. }
    
19. 
    
20. stargazer.keepstats<-function(call="") {
    
21. command<-gsub("\\)$",", table.layout='s'\\)",command)
    
22. call<-eval(parse(text=command))
    
23. row1<-gsub("(.*)\\\\\\\\\\[-1\\.8ex\\](.*)(\\\\end\\{tabular\\})(.*)","\\2",paste(call,collapse=" "))
    
24. row1
    
25. }
    
26. stargazer.begintable<-function(call="") {
    
27. command<-gsub("\\)$",", table.layout='m'\\)",command)
    
28. call<-eval(parse(text=command))
    
29. row1<-paste("\\begin{tabular}",gsub("(.*)\\\\begin\\{tabular\\}(.*)(\\\\end\\{tabular\\})(.*)","\\2",paste(call,collapse="\n")),sep="")
    
30. row1
    
31. }
    
32. stargazer.varlabels<-function(call="") {
    
33. command<-gsub("\\)$",", table.layout='d'\\)",command)
    
34. call<-eval(parse(text=command))
    
35. row1<-paste("\\begin{tabular}",gsub("(.*)\\\\begin\\{tabular\\}(.*)(\\\\end\\{tabular\\})(.*)","\\2",paste(call,collapse="\n")),sep="")
    
36. row1
    
37. }
    
38. 
    
39. stargazer.keepcollabels<-function(call="") {
    
40. command<-gsub("\\)$",", table.layout='c'\\)",command)
    
41. call<-eval(parse(text=command))
    
42. row1<-gsub("(.*)\\\\\\\\\\[-1\\.8ex\\](.*)(\\\\end\\{tabular\\})(.*)","\\2",paste(call,collapse=" "))
    
43. row1
    
44. }
    
45. 
    
46. stargazer.keepomit<-function(call="") {
    
47. command<-gsub("\\)$",", table.layout='o'\\)",command)
    
48. call<-eval(parse(text=command))
    
49. row1<-gsub("(.*)\\\\\\\\\\[-1\\.8ex\\](.*)(\\\\end\\{tabular\\})(.*)","\\2",paste(call,collapse="\n"))
    
50. row1
    
51. }It easiest to see how you can use these functions to construct stacked regression output by giving a simple example.###the global command to be passed to the hacky ###functions that extract the individual bits
    
52. ## OLS is a list of OLS results from running the
    
53. ## lfe command
    
54. command<-'stargazer(OLS , keep=c("post08shale:directutilityshare","post08shale"), covariate.labels=c("Energy Intensity x Shale","Shale"), header=FALSE, out.header=FALSE, keep.stat=c("n","adj.rsq"))'
    
55. 
    
56. begintable<-stargazer.begintable(command)
    
57. ###some multicolumn to combine
    
58. collcombine<-c("& \\multicolumn{4}{c}{Tradable Goods Sector Only} & \\multicolumn{3}{c}{Additional Sectors} & \\cmidrule(lr){2-5} \\cmidrule(lr){6-8}")
    
59. collabel<-stargazer.keepcollabels(command)
    
60. colnums<-stargazer.keepcolnums(command)
    
61. ##plain OLS row
    
62. row1<-stargazer.keeprow(command)
    
63. ##the stats part for the OLS (number of Obs, R2)
    
64. stats<-stargazer.keepstats(command)
    
65. ##the rows for the Fixed effect indicators
    
66. omitted<-stargazer.keepomit(command)
    
67. 
    
68. ##the IV command passing a list of IV results in ## IV object
    
69. command<-'stargazer(IV , keep=c("post08anywell:directutilityshare","post08anywell"), covariate.labels=c("Energy Intensity x Anywell","Anywell"), header=FALSE, out.header=FALSE, keep.stat=c("n","adj.rsq"))'
    
70. 
    
71. ##IV row
    
72. row2<-stargazer.keeprow(command)
    
73. footer<-c("\\end{tabular}\n" )
    
74. 
    
75. ###now combine all the items
    
76. cat(begintable,collcombine,collabel,colnums,"\\hline \\\\\\emph{Reduced Form} \\\\",row1,""\\hline \\\\\\emph{Reduced Form} \\\\",row2,stats,"\\hline\\hline",footer, file="Draft/tables/energyintensity.tex", sep="\n")

复制代码

I know the solution is a bit hacky, but it works and does the trick.

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：Regressions regression regressio regress tables particular directly correct please impact