Transcription of Stata: Visualizing Regression Models Using coefplot
1 stata : Visualizing Regression Models Using coefplotPartiallybased on Ben Jann s June 2014 presentation at the 12thGerman stata Users Group meeting in Hamburg, Germany: A new command for plotting Regression coefficients and other estimates on other materials from coefplot website: KoffmanOffice of Population Research (OPR)Princeton UniversityMay 9, 20171 Workshop OutlineMotivationIntroductioncoefplotcom mand Basic usage single model multiple Models subgraphs Labels confidence intervals2 Motivation Regression results are often presented in tablesdiabetesfemale ( )age **( )bmi **( )region==NE ( )region==MW ( )region==S **( )** p< , ** p< , * p< Regression results are often presented in tables however, displaying results graphically can be much more effective:easier to see and remember patterns and trendsfemaleagebmiregion==NEregion==MWre gion== ratio: diabetesdiabetesfemale ( )age **( )bmi **( )region==NE ( )region==MW ( )region==S **( )** p< , ** p< , * p< Regression results are often presented in tables however, displaying results graphically can be much more effective:easier to see and remember patterns and trendsOUTCOME diabetes SAMPLE Rural Non Ruralfemale ( ) ( )age ** **( ) ( )bmi ** **( ) ( )region==NE ( ) ( )region==MW ( ) ( )region==S **( ) ( )** p< , ** p< , * p< ratio: diabetesRuralNon-Rural5 MotivationFor many more examples,see:Kastellec, Jonathan P.
2 , Eduardo L. Leoni (2007). Using Graphs Instead of Tables in Political Science. Perspectives on Politics 5(4): 755 vs Graphstablesuseful for seeing individual, precise values graphsallow reader to easily: see patterns within a set of Regression results make comparisonsparticularly useful when presenting results: from several Models Using different samples7coefplot IntroductionStata command for graphing results of stata estimation commands user written author: Ben Jann, University of Berndefault behavior plots markers for coefficients and horizontal spikes for confidence intervalsfeatures results from multiple Models can be displayed on a single graph results from multiple Models can be displayed on multiple subgraphs labels can be applied to coefficients, groups of coefficients, and subgraphs sub headings can be inserted to structure the display confidence intervals can show multiple levels8coefplot Resourcescoefplot website.
3 Journal article by Ben Jann Volume 14, Number 4 Plotting Regression coefficients and other Paper by Ben JannCurrent version February 9, 2017, First version August 25, 2013 Plotting Regression coefficients and other estimates in 2013 coefplotAuthor request: Thanks for citing coefplot in your work in one of the following ways:Jann, Ben (2014). Plotting Regression coefficients and other estimates. The stata Journal 14(4): 708 , Ben (2014). Plotting Regression coefficients and other estimates in stata . University of Bern Social Sciences Working Papers No. 1. Available from , Ben (2013). coefplot : stata module to plot Regression coefficients and other results. Available from 10 Install User Written coefplot CommandIn stata , run command:ssc install coefplot , replaceTo view the help file:help coefplot11A Single Modellogit diabetes female age bmi reg1 reg2 reg3 reg4, or-------------------------------------- ---------------------------------------- diabetes | Odds Ratio Std.
4 Err. z P>|z| [95% Conf. interval ]-------------+----------------- ---------------------------------------- -------female | .1020453 .8834407 | .0040456 | .0088939 | .1555003 .8151347 | .1475434 .8367247 | .1696527 | 1 (omitted)_cons | .0002576 .0000868 .000133 .0004987-------------------------------- ---------------------------------------- ------estimates store fem_age_bmi_reg12webuse nhanes2, cleartab region, gen(reg)Basic Usage: Plotting a Single Modelcoefplot fem_age_bmi_reg1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==S_cons- 10-8-6-4-20clearly, notodds ratios! 13 Basic Usage: Plotting a Single Modelcoefplot fem_age_bmi_reg, eformodds ratio s, but don t wantto display constant1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion== Usage: Plotting a Single Modelcoefplot fem_age_bmi_reg, eformdrop(_cons)1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion== better, but uselinear or log scale todisplay odds ratio s ?
5 ?15 Basic Usage: Plotting a Single Modelcoefplot fem_age_bmi_reg, eform drop(_cons) xscale(log)add a lineat x == 1 ?1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion== Usage: Plotting a Single Model#delimit ; coefplot fem_age_bmi_reg, eform drop(_cons) xscale(log) xline(1);1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion== omittedodds ratio?17 Basic Usage: Plotting a Single Model#delimit ; coefplot fem_age_bmi_reg, eform drop(_cons) xscale(log) xline(1) omitted;1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == titles and text?18 Basic Usage: Plotting a Single Model#delimit ; coefplot fem_age_bmi_reg, eform drop(_cons) xscale(log) xline(1) omittedtitle("Diabetes") xtitle("odds ratio")text( "N=10,349", box fcolor(white) lcolor(black))note("Note. Data source: nhanes2", span)graphregion(fcolor(white)); N=10,3491=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == ratioNote.
6 Data source: nhanes2 Diabetes19 Basic Usage: Plotting Multiple Modelsquietly logit diabetes female age bmi reg1 reg2 reg3 reg4 if rural == 1, orestimates store ruralquietly logit diabetes female age bmi reg1 reg2 reg3 reg4 if rural == 0, orestimates store nonrural#delimit ; coefplot rural nonrural, eform drop(_cons) xscale(log range(.5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) ;1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == Usage: Plotting Multiple Models #delimit ; coefplot (rural, mcolor(green) ciopts(lcolor(green))) (nonrural, mcolor(orange) ciopts(lcolor(orange))), eform drop(_cons) xscale(log range(.5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) msymbol(D);1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == Usage: Plotting Multiple Models #delimit ; coefplot (rural, offset(.08)) (nonrural, offset( )),eform drop(_cons) xscale(log range(.))
7 5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) ;1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == Usage: Plotting Multiple Models23#delimit ; coefplot (rural, offset(.08)) (nonrural, offset( ) mcolor(white) ciopts(lcolor(white))),eform drop(_cons) xscale(log range(.5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash))xsize( ) ysize( );1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == Usage: Plotting Multiple Models24#delimit ; coefplot (rural, offset(.08)) (nonrural, offset( )),eform drop(_cons) xscale(log range(.5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash))xsize( ) ysize( );1=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == Usage: Subgraphs25#delimit ;quietly logit diabetes female age bmi reg1 reg2 reg3 reg4 if rural == 1, or;estimates store rural;quietly logit diabetes female age bmi reg1 reg2 reg3 reg4 if rural == 0, or;estimates store nonrural;quietly logit heartatkfemale age bmi reg1 reg2 reg3 reg4if rural == 1, or;estimates store hrural;quietly logit heartatkfemale age bmi reg1 reg2 reg3 reg4if rural == 0, or;estimates store hnonrural;;Basic Usage: Subgraphs261=female, 0=maleage in yearsBody Mass Index (BMI)region==NEregion==MWregion==Sregion == attackruralnonrural#delimit ; coefplot (rural, offset(.
8 08)) (nonrural, offset( )), bylabel(diabetes) ||(hrural, offset(.08)) (hnonrural, offset( )), bylabel(heart attack) ||,eform drop(_cons) omitted xline(1, lcolor(black) lwidth(thin)) xscale(log range(.25 2)) xlabel(.25 ".25" .5 ".5" 1 "1" 2 "2") ;Labels27#delimit ; coefplot (rural, label(Rural Sample) offset(.08)) (nonrural, label(Non-Rural Sample) offset( )),coeflabels(female="female" age="age (years)" bmi="bmi" reg1="region: NE" reg2="region: MW" reg3="region: S" reg4="(ref) region: W", notick labgap(2))eform drop(_cons) xscale(log range(.5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) ;femaleage (years)bmiregion: NEregion: MWregion: S(ref) region: SampleNon-Rural SampleLabels28#delimit ; coefplot (rural, label(Rural Sample) offset(.08)) (nonrural, label(Non-Rural Sample) offset( )),coeflabels(female="female" age="age (years)" bmi="bmi" reg1="region: NE" reg2="region: MW" reg3="region: S" reg4="region: W (reference)", notickwrap(12) labsize(medlarge) labcolor(purple) labgap(2))eform drop(_cons) xscale(log range(.
9 5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) ;femaleage (years)bmiregion: NEregion: MWregion: Sregion: W(reference). SampleNon-Rural SampleLabels29#delimit ; coefplot (rural, label(Rural Sample) offset(.14)) (nonrural, label(Non-Rural Sample) offset( )),coeflabels(female="female" age="age (years)" bmi="bmi" reg1="North East" reg2="Mid West" reg3="South" reg4="(reference cat) West", notick labsize(medlarge) labcolor(purple) labgap(2))headings(female="{bf:Demograph ics}" bmi="{bf:Health}" reg1="{bf:Region}", labcolor(blue))eform drop(_cons) xscale(log range(.5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) ;femaleage (years)bmiNorth EastMid WestSouth(reference cat) SampleNon-Rural SampleLabels30#delimit ; coefplot (rural, label(Rural Sample) offset(.14)) (nonrural, label(Non-Rural Sample) offset( )),coeflabels(female="female" age="age (years)" bmi="bmi" reg1="North East" reg2="Mid West" reg3="South" reg4="(reference cat) West", notick labsize(medlarge) labcolor(purple) labgap(2))groups(female age bmi="{bf:Demographics}" reg*="{bf:Regions}")eform drop(_cons) xscale(log range(.
10 5 2)) omittedxline(1, lcolor(black) lwidth(thin) lpattern(dash)) ;DemographicsRegionsfemaleage (years)bmiNorth EastMid WestSouth(reference cat) SampleNon-Rural SampleLabels: Markers31quietly logit diabetes female age bmi reg1 reg2 reg3 reg4, orestimates store fem_age_bmi_reg#delimit ; coefplot fem_age_bmi_reg, eform drop(_cons) xscale(log) xline(1, lwidth(vthin)) omitted yscale(off) mlabsize(medium)graphregion(fcolor(white ))mlabels(female=12 "female" age=12 "age (in years)" bmi=12 "bmi"reg1=12 "Northeast region" reg2=12 "Midwest region" reg3=12 "South region" reg4=12 "West region (reference category)");femaleage (in years)bmiNortheast regionMidwest regionSouth regionWest region (reference category). : Markers32#delimit ; coefplot (rural, label(Rural Sample) offset(.1) mlabels(female=12 "female" age=12 "age (in years)" bmi=12 "bmi"reg1=12 "Northeast region" reg2=12 "Midwest region" reg3=12 "South region" reg4=12 "West region (reference category)") mlabsize(medium) mlabcolor(black))(nonrural, label(Non-Rural Sample) offset( )),eform drop(_cons) xscale(log range(.
