STATA ‘tricks’: Creating professional graphs to enhance communication of statistical results

Dr. Paul Sirma – Economist, AIR

Prof. Ashu Handa, AIR Institute Fellow and Kenan Eminent Professor of Public Policy, University of North Carolina

Session II: April 2, 2025

Outline for today

​

• Results
• Table of regression output (required)
• Summary graphs of coefficients
• Plot key coefficients from one regression
• Plot key coefficients from the same regression (Y) for different sub-groups
• Plot the same coefficient for different outcomes (Y)

​

Introduction

• Package needed: coefplot
• Displaying regression coefficients and more
• Author: Ben Jann
• Resources: https://repec.sowi.unibe.ch/stata/coefplot/

​

• Installing coefplot package in your computer
• ssc install coefplot, replace

​

• To understand the syntax of coefplot
• Call for help in stata: help coefplot

​

Recap from last session

Rate of not in employment, enrollment or training (NEET) is increasing in urban areas over time among young men in Ghana

The gap is widening among younger men

Goal: graphical representation of the regression output

Urban: �reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural ==0, r

------------------------------------------------------------------------------

| Robust

neet | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | .0205034 .0215487 0.95 0.341 -.0217556 .0627625

year2014 | .0376215 .0222254 1.69 0.091 -.0059646 .0812077

age_2 | .0685399 .0317051 2.16 0.031 .0063631 .1307168

age_3 | .1212137 .0353165 3.43 0.001 .0519545 .1904728

age_4 | .1198266 .0333794 3.59 0.000 .0543662 .1852869

age_5 | .2416338 .0382977 6.31 0.000 .1665282 .3167395

age_6 | .1992317 .038176 5.22 0.000 .1243649 .2740986

age_7 | .2334269 .0411552 5.67 0.000 .1527174 .3141364

age_8 | .1142619 .0368932 3.10 0.002 .0419106 .1866132

age_9 | .0528976 .0332896 1.59 0.112 -.0123866 .1181817

age_10 | .0897838 .0357166 2.51 0.012 .0197399 .1598276

_cons | .0809203 .0231158 3.50 0.000 .0355879 .1262528

------------------------------------------------------------------------------

Rural: �reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural ==1, r

------------------------------------------------------------------------------

| Robust

neet | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | -.0132159 .01463 -0.90 0.366 -.0419025 .0154707

year2014 | -.020363 .0149014 -1.37 0.172 -.0495817 .0088558

age_2 | .0510082 .021771 2.34 0.019 .0083196 .0936967

age_3 | .0277575 .0223311 1.24 0.214 -.0160294 .0715444

age_4 | .0975914 .0234785 4.16 0.000 .0515548 .1436281

age_5 | .068804 .0251296 2.74 0.006 .0195299 .1180781

age_6 | .0513173 .0246995 2.08 0.038 .0028864 .0997482

age_7 | .0190586 .0245162 0.78 0.437 -.0290128 .0671299

age_8 | .0046135 .0240916 0.19 0.848 -.0426253 .0518523

age_9 | .0289109 .0266241 1.09 0.278 -.0232937 .0811156

age_10 | .0100647 .0247859 0.41 0.685 -.0385356 .0586649

_cons | .0891896 .0151379 5.89 0.000 .0595072 .118872

-----------------------------------------------------------------------------

Plotting coefficients from single regression

Urban: �reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural ==0, r

------------------------------------------------------------------------------

| Robust

neet | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | .0205034 .0215487 0.95 0.341 -.0217556 .0627625

year2014 | .0376215 .0222254 1.69 0.091 -.0059646 .0812077

age_2 | .0685399 .0317051 2.16 0.031 .0063631 .1307168

age_3 | .1212137 .0353165 3.43 0.001 .0519545 .1904728

age_4 | .1198266 .0333794 3.59 0.000 .0543662 .1852869

age_5 | .2416338 .0382977 6.31 0.000 .1665282 .3167395

age_6 | .1992317 .038176 5.22 0.000 .1243649 .2740986

age_7 | .2334269 .0411552 5.67 0.000 .1527174 .3141364

age_8 | .1142619 .0368932 3.10 0.002 .0419106 .1866132

age_9 | .0528976 .0332896 1.59 0.112 -.0123866 .1181817

age_10 | .0897838 .0357166 2.51 0.012 .0197399 .1598276

_cons | .0809203 .0231158 3.50 0.000 .0355879 .1262528

------------------------------------------------------------------------------

coefplot

coefplot,mlabel

coefplot , mlabel mlabposition(12)

coefplot , mlabel mlabposition(12) format(%9.2f)

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0)

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) drop(_cons)

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014)

What’s with the horizontal lines on the coefficients?

help coefplot

“levels(numlist) sets the level(s), as percentages, for confidence intervals. Specified values may be between 10.00 and 99.99 and can have at most two digits after the decimal point. The default is levels(95) or as set by set level. If multiple values are specified, multiple confidence intervals are plotted. For example, type levels(99.9 99 95) to plot the 99.9%, 99%, and 95% confidence intervals...

​

ci(spec) specifies the source from which to collect confidence intervals. Default is to compute confidence intervals for the levels specified in levels() using variances/standard errors (and, possibly, degrees of freedom). The ci() option is useful to plot confidence intervals that have been provided by the estimation command”

​

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014) ci(99 95 90)

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014) ci(99 95 90) ciopts(recast(rspike rcap rcap))

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014) ci(99 95 90) ciopts(recast(rcap ..))

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014) ci(99 95 90) ciopts(recast(rcap ..) lc(black ..) lw(medium..))

Back to our default 95% CIs

Urban: coefplot , mlabel mlabposition(12) format(%9.2f) xline(0)

We can run the same regression analysis for rural areas

Rural: �reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural ==1, r

------------------------------------------------------------------------------

| Robust

neet | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | -.0132159 .01463 -0.90 0.366 -.0419025 .0154707

year2014 | -.020363 .0149014 -1.37 0.172 -.0495817 .0088558

age_2 | .0510082 .021771 2.34 0.019 .0083196 .0936967

age_3 | .0277575 .0223311 1.24 0.214 -.0160294 .0715444

age_4 | .0975914 .0234785 4.16 0.000 .0515548 .1436281

age_5 | .068804 .0251296 2.74 0.006 .0195299 .1180781

age_6 | .0513173 .0246995 2.08 0.038 .0028864 .0997482

age_7 | .0190586 .0245162 0.78 0.437 -.0290128 .0671299

age_8 | .0046135 .0240916 0.19 0.848 -.0426253 .0518523

age_9 | .0289109 .0266241 1.09 0.278 -.0232937 .0811156

age_10 | .0100647 .0247859 0.41 0.685 -.0385356 .0586649

_cons | .0891896 .0151379 5.89 0.000 .0595072 .118872

-----------------------------------------------------------------------------

coefplot , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014)

Plot rural and urban models on the same graph

One outcome variable (neet)

Two subgroups (rural and urban)

reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==0, r��reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r

Urban

Rural

Storing estimation results

• After estimating a model (regression/summary statistics), you can store the estimated results and access them later
• Stata command: estimates store NAME
• For example:
• reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==0, r
• estimates store urban

​

• reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r
• estimates store rural

​

• Storing estimates uses up your Stata memory space. Delete your stored estimates when you are done using them to free up space using
• estimates clear

​

​

​

coefplot (urban) (rural)

coefplot (urban) (rural) , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014)

coefplot (urban, label("NEET in urban")) (rural, label(“NEET in rural")), mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014)

coefplot (urban, label("NEET in urban") msymbol(triangle)) (rural, label("NEET in rural") msymbol(square)), mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014)

coefplot (urban, label("NEET in urban") msymbol(triangle)) (rural, label("NEET in rural") msymbol(square)), mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008 year2014) ciopts(lc(black))

Multiple outcomes with the same coefficients

Two outcomes (NEET and employment)

• NEET (0/1) indicator
• = 1if NOT in employment, education or training
• =0 if in employment, education or training

​

• offfarm (0/1) indicator that is
• = 1 if employed outside the farm and
• =0 if employed on the farm

One coefficient

• NEET Rural:
• reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r

​

• Off-farm Rural:
• reg offfarm year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r

NEET Rural: reg neet year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r

------------------------------------------------------------------------------

| Robust

neet | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | -.0132159 .01463 -0.90 0.366 -.0419025 .0154707

year2014 | -.020363 .0149014 -1.37 0.172 -.0495817 .0088558

age_2 | .0510082 .021771 2.34 0.019 .0083196 .0936967

age_3 | .0277575 .0223311 1.24 0.214 -.0160294 .0715444

age_4 | .0975914 .0234785 4.16 0.000 .0515548 .1436281

age_5 | .068804 .0251296 2.74 0.006 .0195299 .1180781

age_6 | .0513173 .0246995 2.08 0.038 .0028864 .0997482

age_7 | .0190586 .0245162 0.78 0.437 -.0290128 .0671299

age_8 | .0046135 .0240916 0.19 0.848 -.0426253 .0518523

age_9 | .0289109 .0266241 1.09 0.278 -.0232937 .0811156

age_10 | .0100647 .0247859 0.41 0.685 -.0385356 .0586649

_cons | .0891896 .0151379 5.89 0.000 .0595072 .118872

------------------------------------------------------------------------------

​

estimates store neet

Off-farm Rural: reg offfarm year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r

​

​

------------------------------------------------------------------------------

| Robust

offfarm | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | .1412723 .0265845 5.31 0.000 .089129 .1934157

year2014 | .1353617 .0264465 5.12 0.000 .083489 .1872343

age_2 | .0492852 .0482447 1.02 0.307 -.0453431 .1439135

age_3 | .0040602 .0483476 0.08 0.933 -.0907698 .0988902

age_4 | .0801057 .0453404 1.77 0.077 -.008826 .1690373

age_5 | .1215867 .0494473 2.46 0.014 .0245997 .2185737

age_6 | .0211714 .0448207 0.47 0.637 -.066741 .1090837

age_7 | .1751065 .0510227 3.43 0.001 .0750295 .2751835

age_8 | .1366944 .0496191 2.75 0.006 .0393704 .2340185

age_9 | .220034 .051742 4.25 0.000 .1185461 .3215218

age_10 | .1901072 .0506619 3.75 0.000 .0907379 .2894766

_cons | .106935 .0356311 3.00 0.003 .0370474 .1768226

------------------------------------------------------------------------------

estimates store offfarm

coefplot (neet ) (offfarm), mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008)

coefplot (neet offfarm ) , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008)

coefplot (neet offfarm), mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008) asequation

coefplot (neet offfarm), mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008) asequation swapname

Adding a third outcome

reg employed year2008 year2014 age_2 age_3 age_4 age_5 age_6 age_7 age_8 age_9 age_10 if rural==1, r

​

------------------------------------------------------------------------------

| Robust

employed | Coefficient std. err. t P>|t| [95% conf. interval]

-------------+----------------------------------------------------------------

year2008 | .0748356 .0205892 3.63 0.000 .0344643 .1152068

year2014 | .1307284 .0210776 6.20 0.000 .0893993 .1720574

age_2 | .0301977 .0338482 0.89 0.372 -.036172 .0965674

age_3 | .0840416 .0368515 2.28 0.023 .0117831 .1563

age_4 | .1625156 .0345181 4.71 0.000 .0948324 .2301988

age_5 | .2378392 .0373011 6.38 0.000 .164699 .3109793

age_6 | .3596835 .0365082 9.85 0.000 .2880981 .4312689

age_7 | .3820857 .0386875 9.88 0.000 .3062272 .4579442

age_8 | .4787038 .0357186 13.40 0.000 .4086667 .5487409

age_9 | .4872589 .0362585 13.44 0.000 .4161632 .5583546

age_10 | .5062429 .034975 14.47 0.000 .4376639 .5748219

_cons | .2901619 .0257977 11.25 0.000 .2395776 .3407462

------------------------------------------------------------------------------

​

estimate store employed

​

​

Using options ‘asequation’ and ‘swapnames’ produce nicer graphs when you have multiple outcomes

coefplot (neet offfarm employed ) , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008)

coefplot neet offfarm employed , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008)

Nicer graph by swapping equation names with coefficient names

coefplot (neet offfarm employed ) , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008) asequation

coefplot (neet offfarm employed ) , mlabel mlabposition(12) format(%9.2f) xline(0) keep(year2008) asequation swapnames