Stata备忘录
1. 画图
（1）时间趋势图
label var year "年份"
label var per "制造业增加值比重[左轴]"
label var tjj "工业增加值比重[右轴]"
graph twoway (connect per year ,yaxis(1) color(black) ) ///
(connect tjj year ,yaxis(2) color(black) lpattern(dash) ) ///
, graphregion(color(white)) xlabel(2003(2)2019) ///
ytitle("世界银行制造业增加值比重(%)",axis(1) height(5)) ///
ytitle("国家统计局工业增加值比重(%)",axis(2) height(5)) ///
note(数据来源：World Bank Open Data、国家统计局) xline(2011)
等价命令
tw (connect value1819 season , ///
lcolor(black) lpattern(dash) msymbol(O) mlcolor(gs5) mfcolor(gs12)) ///
(connect value2020 season , ///
lcolor(black) lpattern(solid) msymbol(S) mlcolor(gs5) mfcolor(gs12)) ///
,graphregion(color(white)) ///
legend(label(1 "18-19年平均") label(2 "2020年") ) ///
xlabel(1 "第一季度" 2 "第二季度" 3 "第三季度" 4 "第四季度" ,labsize(small) )
label var year "年份"
tw bar mR1 year,yaxis(2) bc(balck) sort barwidth(0.9) fintensity(inten0) ///
ylabel(0(2000)6000, axis(2)) ///
xlabel(2014(1)2021)|| ///
connect percent_R year,yaxis(1) lc(black) lp(dash) mc(blace) ///
ylabel(0.5 "50%" 0.6 "60%" 0.7 "70%" 0.8 "80%" ,axis(1)) ||, ///
graphregion(color(white) ) ///
bgcolor(white) ///
title("中国数字内容企业(游戏)收入金额及占全球市场比重", c(black) size(*0.8)) ///
ytitle("占比（%）",axis(1) height(7)) ///
ytitle("收入额（百万美元）",axis(2) height(5)) ///
legend(label(1 "中国数字内容企业(游戏)收入占全球市场比重") label(2 "中国数字内容企业(游戏)收入金额") ) ///
legend(size(small) col(1)) ///
note("数据源自：app annie")
graph save "Graph" "$path\output\playdata_1_percent_and_value_of_Chinese_Apps_Export.gph",replace
use hs_adj_year_PQV_2000_2015.dta,clear
use hs_adj_year_PQV_2000_2015_cregime10,clear
merge m:1 hs_adj using equipment
replace BEC=4 if BEC==1 & equipment!=1
destring hs_adj,replace
reghdfe lnV i.year if year!=2006 & BEC==2, a(hs)
est store result_accessories
reghdfe lnV i.year if year!=2006 & BEC==4, a(hs)
est store result_equipment
reghdfe lnV i.year if year!=2006 & BEC==0, a(hs)
est store result_noncapital
#d ;
coefplot
(result_accessories,c(l) label("accessories") lp(dash) lc(black) mc(black) ms(smcircle_hollow) offset(-0.07))
(result_equipment ,c(l) label("equipment") lp(solid) lc(black) mc(black) ms(smcircle_hollow))
(result_noncapital,c(l) label("noncapital") lp(dot) lc(black) mc(black) ms(smcircle_hollow) offset(0.07))
, vertical
drop(_cons) byopts(xrescale)
xlabel(1 "2001" 3"2003" 5"2005" 6"2007" 8"2009" 10"2011" 12"2013" 14"2015")
graphregion(color(white))
legend(size(small) col(3))
;
#d cr
字体大小 option
| 字体大小option | description |
|---|---|
| zero | no size whatsoever, vanishingly small |
| minuscule | smallest |
| quarter_tiny | |
| third_tiny | |
| half_tiny | |
| tiny | |
| vsmall | |
| small | |
| medsmall | |
| medium | |
| medlarge | |
| large | |
| vlarge | |
| huge | |
| vhuge | largest |
| tenth | one-tenth the size of the graph |
| quarter | one-fourth the size of the graph |
| third | one-third the size of the graph |
| half | one-half the size of the graph |
| full | text the size of the graph |
| size | any size you want |
节点样式 eg: msymbol(O) mlcolor(gs5) mfcolor(gs12)
| symbolstyle | Synonym(if any) | Description |
|---|---|---|
| circle | O | solid |
| diamond | D | solid |
| triangle | T | solid |
| square | S | solid |
| plus | + | |
| X | X | |
| arrowf | A | filled arrow head |
| arrow | a | |
| pipe | ||
| V | V | |
| smcircle | o | solid |
| smdiamond | d | solid |
| smsquare | s | solid |
| smtriangle | t | solid |
| smplus | ||
| smx | x | |
| smv | v | |
| circle_hollow | Oh | hollow |
| diamond_hollow | Dh | hollow |
| triangle_hollow | Th | hollow |
| square_hollow | Sh | hollow |
| smcircle_hollow | oh | hollow |
| smdiamond_hollow | dh | hollow |
| smtriangle_hollow | th | hollow |
| smsquare_hollow | sh | hollow |
| point | p | a small dot |
| none | i | a symbol that is invisible |
线样式
| linepatternstyle | Description |
|---|---|
| solid | solid line |
| dash | dashed line |
| dot | dotted line |
| dash_dot | |
| shortdash | |
| shortdash_dot | |
| longdash | |
| longdash_dot | |
| blank | invisible line |
| formula | e.g.,-. or --… etc. |
| A formula is composed of any combination of | |
| l | solid line |
| _ | (underscore) a long dash |
| - | (hyphen) a medium dash |
| . | short dash (almost a dot) |
| # | small amount of blank space |
颜色
| black | edkblue | gs12 | lime | orange |
|---|---|---|---|---|
| blue | eggshell | gs13 | ltblue | orange_red |
| bluishgray | eltblue | gs14 | ltbluishgray | pink |
| bluishgray8 | eltgreen | gs15 | ltbluishgray8 | purple |
| brown | emerald | gs16 | ltkhaki | red |
| chocolate | emidblue | gs2 | magenta | sand |
| cranberry | erose | gs3 | maroon | sandb |
| cyan | forest_green | gs4 | midblue | sienna |
| dimgray | gold | gs5 | midgreen | stone |
| dkgreen | gray | gs6 | mint | sunflowerlime |
| dknavy | green | gs7 | navy | teal |
| dkorange | gs0 | gs8 | navy8 | white |
| ebblue | gs1 | gs9 | none | yellow |
| ebg | gs10 | khaki | olive | |
| edkbg | gs11 | lavender | olive_teal |
（2）柱状图
#delimit ;
graph bar cn_wzje_80 cn_wzje_81 if wzlx==0, over(sec) bargap(-30)
ytitle("吸引外资金额")
legend( label(1 "08年前") label(2 "08年后") )
title("各行业吸引外资2008年前后对比")
subtitle("中国中西部服务业")
note("中国中西部服务业") ;
#delimit cr
（3）散点图
tw (scatter delta_v2_v3 delta_v1_v2 if delta_v1_v2>=-0.3& delta_v1_v2 <=2.3 ///
&delta_v2_v3>= -0.3&delta_v2_v3<=2.3, ///
mlabel(hy4) mlc(black) mlabc(black) ms(x) mlabs(tiny)) ///
(fun y=x,range(-0.3 2.3)) , ///
xlab(-0.3(0.5)2.3) ylab(-0.3(0.5)2.3) ///
graphregion(color(white)) ///
xline(0,lp(dash) lc(gs10)) ///
yline(0,lp(dash) lc(gs10)) ///
legend(ring(0) pos(5) order(2 "45°线")) ///
ytitle("11-15时段的增速") ///
xtitle("07-11时段的增速")
graph twoway (scatter wzje_CE0308 wzje_CE1419, mlabel(sec) mlabv(sec) ) (function y=x, range(0 0.11)) , ///
title("中国东部地区03-08对比中国东部地区14-19[金额]") ///
ytitle("中国东部03-08") ///
xtitle("中国东部14-19") ///
legend(ring(0) pos(5) order(2 "45°线")) ///
graphregion(color(white))
graph twoway (scatter c_AS c_WD, mlabel(cic03) ) (lfit c_AS c_WD) , ///
title("东盟增速放缓 vs 世界增速放缓") ///
ytitle("东盟") ///
xtitle("世界") ///
legend(ring(0) pos(5) order(2 "拟合")) ///
graphregion(color(white))
（4）bgshade
bgshade ks, shaders(uu9) ///
twoway(connect lamda22 ks if treat==1&ks>=6&ks<=11 || ///
connect lamda22 ks if treat==0&ks>=6&ks<=11 , xlab(6(1)11) ///
title("新冠疫情冲击下企业平均收入变化趋势"))
（5）coefplot
coefplot, levels(90) vertical lcolor(black)mcolor(black) ///
msymbol(circle_hollow) ytitle(估计系数, size(small)) ///
ylabel(, labsize(small) angle(horizontal) nogrid) ///
yline(0, lwidth(vthin)lpattern(solid) lcolor(black)) ///
xtitle(事件发生时间, size(small)) ///
title("(B)企业缴税的平行趋势检验") ///
xlabel(0"." 1"2019s2" 2"2019s3" 3"2019s4" 4"2020s1" 5"2020s2" 6"2020s3" 7"2020s4")
reghdfe lnQ i.Year if elec == 1,a(i.citycode) vce(r)
est store elec_Q_1
reghdfe lnV i.Year if elec == 1,a(i.citycode) vce(r)
est store elec_V_1
reghdfe lnQ i.Year if elec == 0,a(i.citycode) vce(r)
est store elec_Q_0
reghdfe lnV i.Year if elec == 0,a(i.citycode) vce(r)
est store elec_V_0
coefplot (elec_Q_1,label("半导体电子元件相关企业进口数量") offset(0.05) pstyle(p3)) ///
(elec_Q_0 ,label("非半导体电子元件相关企业进口数量") offset(-0.05) pstyle(p4) ), ///
vertical drop(_cons) xline(0) ///
graphregion(color(white)) ///
yline(0) ///
addplot(line @b @at,lp(dash) lwidth(*0.5)) ///
legend(label(1 "半导体电子元件相关企业进口数量") label(2 "非半导体电子元件相关企业进口数量") )
（6）画系数和置信区间
twoway (scatter coef week) ///
(rcap ci_lower ci_upper week, ///
lcolor(black) ///
mcolor(black) ///
lwidth(vthin) ///
lpattern(dash) ///
msymbol(circle_hollow) ///
legend(label(2 "99% CI"))) , ///
yline(0) ///
xtitle("") ///
graphregion(fcolor(white)) ///
title("第X周的系数", size(medium)) ///
name("Coef_all_I", replace)
（7）画直方图
一般使用kdensity
hist year if year>=1400 & year<=2010, freq bin(200) ylabel(0(500)2500) xtitle("Year") xline(1950 1980,lw(thin)) ///
text(1500 1950 "Year=1950", place(w)) text(2000 1980 "Year=1980", place(w))
（8）画桑基图
cd $path\appdata
use Data_games.dta,clear
merge m:1 ParentCompanyName using "$path\data\company_city"
keep if _m == 3
drop _m
gen from = city_code
encode iso3_j,gen(to)
bys from to :egen tR = total(Revenue)
bys from to :egen tD = total(Downloads)
duplicates drop from to ,force
gen x0 = 1
gen x1 = 2
tostring city_code ,gen(city2)
drop if dest == "CHN"
sankey_plot x0 from x1 to, ///
width0(tR) extra xlabel(1 "Source" 2 "Destination", nogrid labsize(small)) ///
colorpalette(economist, opacity(30)) ///
label0(city) label1(iso3_j) ///
labsize(*0.6) labcolor(black) ///
graphregion(color(white)) gap(0.1) ///
title("地级市层面Apps出海流向（按收入额）",color(black) size(*0.8))
graph save "Graph" "$path\output\sankey_R_0228.gph",replace
sankey_plot x0 from x1 to, ///
width0(tD) extra xlabel(1 "Source" 2 "Destination", nogrid labsize(small)) ///
colorpalette(economist, opacity(30)) ///
label0(city) label1(iso3_j) ///
labsize(*0.6) labcolor(black) ///
graphregion(color(white)) gap(0.1) ///
title("地级市层面Apps出海流向（按下载量）",color(black) size(*0.8))
graph save "Graph" "$path\output\sankey_D_0228.gph",replace
（9）气泡图
twoway(scatter mv T_gap_05_00 [fweight=N] if BEC == 0&T_gap_05_00!=0&N !=0,msymbol(Oh) mc(ebblue%40)) ///
(scatter mv T_gap_05_00 [fweight=N] if BEC == 4&T_gap_05_00!=0&N !=0,msymbol(Oh) mc(orange_red%40)) ///
(scatter mv T_gap_05_00 [fweight=N] if BEC == 2&T_gap_05_00!=0&N !=0,msymbol(Oh) mc(green%20)) ///
, legend(label(1 "非资本品") label(2 "equipment") label(3 "accessories") )
2. 处理数据
（1）拓展expand数据
例如：当前数据中 有9523、9524、9525、9526、9527、9528共计6个样本，现在想把这6个样本根据freq进行扩充，变为9523、9524_1、9524_2、9525_1、9525_2、9526_1、9526_2、9527_1、9527_2、9528_1、9528_2这11个样本
| freq | count | value |
|---|---|---|
| 1 | 9523 | 4845.1143 |
| 2 | 9524 | 969.66498 |
| 2 | 9525 | 129.53349 |
| 2 | 9526 | 71284.508 |
| 2 | 9527 | 1038.127 |
| 2 | 9528 | 445877.09 |
count是id的唯一识别码，expandcl函数可以生成freq行相同的样本，并生成一个新的id识别码freq_count
egen count=group(id hs02_6)
expandcl freq,gen(freq_count) cluster(count)
drop freq_count
（2）时间数据
gen R= mdy(month_r,day_r,year_r)
gen week_r = week(R)
gen day_r = day(R)
gen dow_r = dow(R) //返回周几
gen doy_r = doy(R) //返回年内日期
gen yw_r = yw(year_r,week_r)
gen ed = yw - yw_r
//yw ym yq yh分别为年周、年月、年季、年半年
gen period_kb= date(date_u,"YMD")-date(date_kb,"YMD")
（3）常见函数
int(x) //取整，不论后面的小数是什么，只取小数点前的数值
round(x) // 四舍五入取整
round(x, .01) //保留两位小数四舍五入
gen y = sum(x) //求列累积和
egen y = sum(x) //求列总和
egen y = rsum(x y z) //求x+y+z总和
egen y = rowmean(x y z) //求(x+y+z)/3
egen y = rowsd(x y z) //求x y z的方差
egen y = rowmim(x y z) //求x y z的最小值
egen y = rowmax(x y z) //求x y z的最大值
egen y = mean(x) //求列均值
egen y = median(x) //求列中位数
egen y = std(x) //求变异系数，与方差不同
bysort x(y): gen z = y[1] //按照x分组，分组后按照y排序，生成一个新变量z=y的第一个观察值
（4）缩尾处理
foreach v of var DexpoAS4- DlnexpoWD2{
gen `v'_w=`v'
qui su `v',det
replace `v'_w=r(p99) if `v'>r(p99) & `v'<.
replace `v'_w=r(p1) if `v'<r(p1)
}
winsor2 wage, replace cuts(1 99) trim
summary 一个变量之后，可以返回的结果有
r(N) //number of observations
r(mean) //mean
r(skewness) //skewness (detail only)
r(min) //minimum
r(max) //maximum
r(sum_w) //sum of the weights
r(p1) //1st percentile (detail only)
r(p5) //5th percentile (detail only)
r(p10) //10th percentile (detail only)
r(p25) //25th percentile (detail only)
r(p50) //50th percentile (detail only)
r(p75) //75th percentile (detail only)
r(p90) //90th percentile (detail only)
r(p95) //95th percentile (detail only)
r(p99) //99th percentile (detail only)
r(Var) //variance
r(kurtosis) //kurtosis (detail only)
r(sum) //sum of variable
r(sd) //standard deviation
（5）创建文件夹
在project路径下生成一个workspace文件夹
再生成子文件夹储存数据（data）、控制变量（controlvars ）、临时数据（tempdata）、回归结果（outreg）
efolder, cd(D:\stata15\project\workspace)
efolder, cd(D:\stata15\project\workspace sub(data controlvars tempdata outreg)
（6）分组处理数据（bysort的替代方案）
*展示根据highzupu50（族谱）和year分组后的变量drqianfen(死亡率)均值；
collapse (mean) drqianfen, by(highzupu50 year)
（7）定义无缺失的样本
g rsample = !mi(avggrain_fyr) & !mi(nograin_fyr) & !mi(urban_fyr)& !mi(dis_bj_fyr) & !mi(dis_pc_fyr) & !mi(migrants_fyr)& !mi(rice_fyr) & !mi(minor_fyr) & !mi(edu_fyr)
（8）定义Dummy的新替代式（时间range）
*如果yob满足1825≤yob≤1899则pre取值为1,否则pre取值为0。mid、post生成过程类似。
gen pre = inrange(yob, 1825, 1899)
gen mid = inrange(yob, 1899, 1919)
gen post = inrange(yob, 1920, 1960)
（9）快速替换
recode treatyear (1969 = 1) (1979 = 2) (1989 = 3) (1999 = 4) (2009 = 5)
3. 处理字符
（1）替换字符
replace 候选人姓名=subinstr( 候选人姓名, " ", "",. )
（2）捕捉字符中的某些特征
keep if strmatch(city, "*山东*")
gen temp = 1 if strmatch(reporteriso3, "A*")
（3）提取字符，检索特定字符
//从enddate字符1开始取，取4个字符赋给year
gen year = substr(enddate,1,4)
//strpos(s1, s2)返回字符s2在s1中的位置，如果s1中找不到s2，则返回0，将该判断再赋给y
gen y = strpos(s1, s2) != 0
4. 输出结果
（1）常规输出
outreg2 using "E:\mfg\outreg\r2", word append addtext(CountryFE, YES,YearFE, YES)
（2）iv回归输出第一阶段
eststo: xtivreg p_a_w (DexpoCN4_w=dexpo44) ///
i.t c.expr0#t c.Lshare0#t ///
c.lnGDP0#t c.lngdp0#t, fe first vce(cluster c)
eststo: xtreg DexpoCN4_w dexpo44 ///
i.t c.expr0#t c.Lshare0#t ///
c.lnGDP0#t c.lngdp0#t if e(sample)==1 ,fe
cd $path\outreg
outreg2 using "table3", word replace addtext(CityFE, YES,YearFE, YES) keep(dexpo44)
（3）变量描述性统计
*列出inv等变量的样本数、均值、标准差、最小值和最大值。
tabstat inv loginv log_levies ///
logpopl logincome logasset hhsize landpc logmigration logtax logtransfer share_admin ///
postcont postopen secret_ballot proxy_voting moving_ballot ///
, s(N mean sd min max) c(s)
5.矩阵保存结果
mat T1 = J(3,3,.)
reghdfe temp ib1.season if year == 2018 & treat == 1,noa
forvalues i = 1/3{
local j = `i' + 1
mat T1[`i',1] = _b[`j'.season]
}
reghdfe temp ib1.season if year == 2019 & treat == 1,noa
forvalues i = 1/3{
local j = `i' + 1
mat T1[`i',2] = _b[`j'.season]
}
reghdfe temp ib1.season if year == 2020 & treat == 1,noa
forvalues i = 1/3{
mat T1[`i',3] = _b[`=1+`i''.season]
}
svmat T1
6.导入数据（全字符串）
forv i = 2000/2003{
cd E:\Data\EPS工企海关匹配库\origindata
import delimited "工企+海关(`i').csv", stringcols(_all) clear
cd E:\Data\EPS工企海关匹配库
save data`i'.dta,replace
}
7.循环
clear all
set obs 1000
**#** 用forvalues循环对单一变量进行处理
gen id = .
//生成一个变量名为id，代表第几个人，假设一共有50个人
//假设每个人都有20个观测值，代表20年
forvalues i = 1/50 {
local j = `i' - 1 //暂时定义0~49 方便计算
local lower = `j' * 20 +1 //定义下限 1、21、41、61
local upper = `j' * 20 + 20 //定义上限 20、40、60、80
//由此就定义了 1~20 21~40 41~60 ……
replace id = `i' in `lower'/`upper' //给第1~20行，赋值为第1个人
//给第21~40行，赋值为第2个人
}
bys id : gen T = _n + 2000 //对于每个人，都生成一个时间序列
**#** 用forvalues循环对多个变量进行处理
forvalues i = 1/5 {
gen value`i' = .
cap gen e = rnormal()
replace value`i' = e * 10 + `i'
cap drop e
}
// 等价于
gen value6 = .
cap gen e = rnormal()
replace value6 = e * 10 + 6
cap drop e
gen value7 = .
cap gen e = rnormal()
replace value7 = e * 10 + 7
cap drop e
gen value8 = .
cap gen e = rnormal()
replace value8 = e * 10 + 8
cap drop e
gen value9 = .
cap gen e = rnormal()
replace value9 = e * 10 + 9
cap drop e
gen value10 = .
cap gen e = rnormal()
replace value10 = e * 10 + 10
cap drop e
**#** 用while循环对单一变量进行处理
// 只要时间在T=11和T=20之间，就对value1~value10进行 " 乘0.1"的处理
local i = 2010
while `i' < 2020 {
forvalues j = 1/10{
replace value`j' = value`j' * 0.1 if T == `i'
}
local i = `i' + 1
}
**#** 用foreach对变量进行处理
foreach v in value1 value2 value3 value4 value5 {
su `v' ,d
replace `v' = (`v' - r(min)) / (r(max) - r(min))
kdensity `v'
}
// 等价于
su value6,d
replace value6 = (value6 - r(min)) / (r(max) - r(min))
kdensity value6
su value7,d
replace value7 = (value7 - r(min)) / (r(max) - r(min))
kdensity value7
su value8,d
replace value8 = (value8 - r(min)) / (r(max) - r(min))
kdensity value8
su value9,d
replace value9 = (value9 - r(min)) / (r(max) - r(min))
kdensity value9
su value10,d
replace value10 = (value10 - r(min)) / (r(max) - r(min))
kdensity value10
8.将第一行作为 label / varname
* 把第一行作为变量标签
labone,nrow(1)
* 把第一行作为变量名（不保留第一行）
nrow
* 把第一行作为变量名（保留第一行）
nrow,keep