Creating coefplot after linear and instrumental-variable regressions

[Brendan Clark]

Hi,

I'm trying to use lincom to plot the linear combination of coefficients for my linear and instrumental-variable regressions.


Problem is, I'm not sure how to store my output and plot it using coefplot with the respective confidence intervals.

This is my attempt with a snippet of my data:

* Example generated by -dataex-. To install: ssc install dataex
clear
input double re_schedule float(post high_female_manager2 high_female_crew2) str3 store_code float periodo
143 0 0 1 "125" 4
94 1 0 1 "125" 25
136 1 0 1 "125" 35
116 1 0 1 "125" 24
305 1 0 1 "125" 27
142 1 0 1 "125" 16
85 0 0 1 "125" 5
71 1 0 1 "125" 14
53 0 0 1 "125" 6
83 1 0 1 "125" 20
84 1 0 1 "125" 23
83 1 0 1 "125" 13
192 1 0 1 "125" 33
199 1 0 1 "125" 17
64 0 0 1 "125" 7
70 0 0 1 "125" 12
93 0 0 1 "125" 11
205 1 0 1 "125" 18
193 0 0 1 "125" 2
15 0 0 1 "125" 0
170 0 0 1 "125" 1
174 1 0 1 "125" 19
180 1 0 1 "125" 32
59 1 0 1 "125" 15
302 1 0 1 "125" 28
99 1 0 1 "125" 22
220 1 0 1 "125" 30
14 1 0 1 "125" 36
257 1 0 1 "125" 26
60 0 0 1 "125" 10
168 0 0 1 "125" 3
71 1 0 1 "125" 21
40 0 0 1 "125" 9
144 1 0 1 "125" 31
220 1 0 1 "125" 34
57 0 0 1 "125" 8
304 1 0 1 "125" 29
15 1 1 0 "138" 27
17 1 1 0 "138" 29
5 1 1 0 "138" 36
20 1 1 0 "138" 20
57 0 1 0 "138" 4
23 1 1 0 "138" 30
14 0 1 0 "138" 8
25 1 1 0 "138" 28
16 1 1 0 "138" 24
34 0 1 0 "138" 5
17 0 1 0 "138" 10
17 1 1 0 "138" 21
112 1 1 0 "138" 34
42 0 1 0 "138" 16
72 0 1 0 "138" 18
17 0 1 0 "138" 12
46 0 1 0 "138" 1
13 0 1 0 "138" 14
11 0 1 0 "138" 7
9 0 1 0 "138" 11
53 0 1 0 "138" 3
16 0 1 0 "138" 6
16 0 1 0 "138" 15
20 1 1 0 "138" 22
29 1 1 0 "138" 31
66 0 1 0 "138" 17
25 1 1 0 "138" 26
15 1 1 0 "138" 23
24 1 1 0 "138" 25
49 0 1 0 "138" 2
88 1 1 0 "138" 33
14 0 1 0 "138" 9
32 0 1 0 "138" 19
29 1 1 0 "138" 32
4 0 1 0 "138" 0
25 0 1 0 "138" 13
73 1 1 0 "138" 35
44 1 0 1 "AND" 19
35 0 0 1 "AND" 3
35 1 0 1 "AND" 20
31 0 0 1 "AND" 12
33 0 0 1 "AND" 1
29 1 0 1 "AND" 27
30 0 0 1 "AND" 13
17 0 0 1 "AND" 5
22 1 0 1 "AND" 32
29 1 0 1 "AND" 23
38 0 0 1 "AND" 11
23 1 0 1 "AND" 30
17 0 0 1 "AND" 7
43 1 0 1 "AND" 34
22 0 0 1 "AND" 9
2 0 0 1 "AND" 0
56 0 0 1 "AND" 17
15 0 0 1 "AND" 6
30 1 0 1 "AND" 21
34 1 0 1 "AND" 29
14 0 0 1 "AND" 14
18 0 0 1 "AND" 15
28 1 0 1 "AND" 26
30 1 0 1 "AND" 33
15 0 0 1 "AND" 10
43 0 0 1 "AND" 4
end

 reghdfe re_schedule i.post##i.high_female_manager2##i.high_female_crew2, abs(store_code periodo) cluster(store_code)

eststo m1: lincom 1.post
eststo m2: lincom 1.post + 1.post#1.high_female_manager2
eststo m3: lincom 1.post + 1.post#1.high_female_crew2
eststo m4: lincom 1.post + 1.post#1.high_female_manager2 + 1.post#1.high_female_crew2 + 1.post#1.high_female_crew2#1.high_female_manager2

coefplot m1 m2 m3 m4, recast(bar)  drop(_cons) vertical

Could someone tell me what I'm doing wrong and how I can plot these estimates after lincom?

For reference, I have attached an image of what I would like my plot to look like.

Thanks!

[Wouter Wakker]

lincom doesn't store anything in e(), so unfortunately cannot be used with eststo. Furthermore, only one linear combination can be calculated at a time. xlincom (from SSC) overcomes both of these limitations.

ssc install xlincom

reghdfe re_schedule i.post##i.high_female_manager2##i.high_female_crew2, abs(store_code periodo) cluster(store_code)

xlincom (imp = 1.post) ///
        (imp_fm = 1.post + 1.post#1.high_female_manager2) ///
        (imp_fc = 1.post + 1.post#1.high_female_crew2) ///
        (imp_fm_fc = 1.post + 1.post#1.high_female_manager2 + 1.post#1.high_female_crew2 + 1.post#1.high_female_crew2#1.high_female_manager2), post
coefplot, recast(bar) vertical