Multiple Bars with 95% CI bar

[Shisho Jakas]

Hi Statalist,

I always appreciate for your help. I am struggling with the confidence interval bar in the graph.

I can successfully made a graph with the help of statalist yesterday, but the thing is that I need to display confidence interval in the graph.

The graph I made is below with the following code.

reg CasCov_a dNative if TipoComRev<=2 
reg CasCov_b dNative if TipoComRev<=2

gen id=_n
reshape long CasCov_, i(id) j(which) string
replace which= "CasCov_a" if which=="a"
replace which= "CasCov_b" if which=="b"

lab def com 0 "Campensino" 1 "Indigineous"
lab val dNative com

graph bar CasCov_ if TipoComRev<=2, over(dNative) ///
over(which, relabel(1 "Case(all)"  2 "Case(confirmed)")) bgcolor(white) plotregion(fcolor(white)) ///
graphregion(fcolor(white)) asyvars showyvars leg(off) ytitle("")

What I need to show is as follows with the CI bar after the regression. If you can see, there is line for 95% CI.

I have been looking forward the command to show it, but coefplot (ssc) seems fine but I cannot find anything to show multiple variables by group in one chart.

Could you let me know the way to draw the graph I showed?

I attach the data below.

Thank you so much for your help in advance.

* Example generated by -dataex-. To install: ssc install dataex
clear
input double UBIGEO float(CasCov_a CasCov_b dNative TipoComRev)
1601070080 1 0 1 2
1601070007 1 0 1 2
1601070072 1 0 0 1
1601100021 0 0 1 2
1605040002 1 0 0 1
1601070010 1 0 1 2
1605050012 1 0 0 1
1601070057 1 0 0 1
1601070074 1 1 1 2
1601060014 1 0 0 1
1607060009 1 0 1 2
2501010062 1 0 0 1
1601070073 1 0 0 1
1601050053 1 0 0 1
1601070090 1 1 1 2
1601070061 1 0 1 2
2501050091 1 0 0 1
1601070048 1 0 1 2
1601070015 1 0 1 2
1601060061 1 0 0 3
1601070071 1 0 1 2
1601070079 1 0 1 2
1601060013 1 0 0 1
1601070030 1 0 1 2
1601060046 1 0 0 1
1601070086 1 0 1 2
1601070024 1 0 1 2
1601070076 1 0 1 2
2501030036 1 0 1 2
1605060013 1 0 1 2
1601070054 1 0 1 2
1607060064 1 0 1 2
1601070014 1 0 1 2
1601070044 1 0 1 2
1601100016 0 0 1 2
1601070006 1 0 1 2
1601070033 1 0 1 2
2501040034 1 1 1 2
2501010197 0 0 0 1
1606020002 0 0 0 1
1601100013 0 0 1 2
2501050062 1 0 1 2
1601070004 1 0 1 2
2501050030 1 0 0 1
1607060051 1 0 1 2
2502030005 1 1 1 2
2501010200 1 0 0 1
1601060024 1 0 0 1
1605060014 1 0 1 2
1601040017 1 0 0 1
1601070005 1 0 1 2
1601060028 1 0 0 1
2501040037 1 0 1 2
1601060017 1 0 0 1
2501030020 1 0 1 2
2502030015 1 1 0 1
1606050021 1 0 0 1
2501040036 1 0 0 1
1601070187 1 0 1 2
2501040094 1 0 0 1
2501070025 1 1 1 2
2501050021 1 1 1 2
1601060015 1 0 0 1
1605060020 1 0 1 2
1601080031 1 0 0 1
2501040032 1 1 1 2
1601060033 1 0 1 2
2501050011 1 0 0 1
1606020005 1 0 0 1
1601060004 1 0 1 2
2501010084 0 0 1 2
1601100010 0 0 1 2
2501070010 1 0 0 1
1601070053 1 0 1 2
1601060012 1 0 1 2
2501040028 1 0 0 1
1606050029 1 1 0 1
1601060045 0 0 0 1
2502030073 1 1 1 2
2501040018 1 1 0 1
2501010094 1 1 1 2
1601060052 1 0 0 1
1601060020 1 0 0 1
1601070088 1 0 0 1
1601050052 1 0 0 1
1601060060 1 0 0 1
1601070002 1 0 1 2
2501040057 0 0 0 1
2501040051 1 0 0 1
1601070067 1 0 0 1
1605060018 1 0 1 2
1601070083 1 0 1 2
2501040015 1 . 0 1
2501030009 1 1 1 2
1601060026 1 0 0 1
2501040025 1 1 1 2
1601080036 1 0 0 1
1601070034 1 0 1 2
2501040048 0 0 0 1
1601080060 1 0 1 2
end
label values CasCov_a CasCov_a_label
label def CasCov_a_label 0 "No", modify
label def CasCov_a_label 1 "Si", modify
label values CasCov_b CasCov_b_label
label def CasCov_b_label 0 "No", modify
label def CasCov_b_label 1 "Si", modify
label values TipoComRev L_TipoCom
label def L_TipoCom 1 "Campesino", modify
label def L_TipoCom 2 "Nativa", modify
label def L_TipoCom 3 "Colono", modify

[Nick Cox]

Your code does absolutely nothing to pick up on any regression results. I wouldn't be surprised if you need to run coefplot after each model and then combine the graphs, but you should get better answers.

[Shisho Jakas]

Dear Nick,

I appreciate that you pointed out the issue. Could you give me a hint to me about how to pick up the regression result and display the CI bar in the graph without using coefplot? Is there any way to do so?

[Shisho Jakas]

With coefplot, I got this for CosCav_a and CosCav_b. This is not close to what I want to get..

proportion CasCov_a if dNative==0 & TipoComRev<=2
estimates store nativea

proportion CasCov_a if dNative==1 & TipoComRev<=2
estimates store campensinoa

coefplot nativea campensinoa, vertical recast(bar) barwidth(0.3) fcolor(*.5) ///
 ciopts(recast(rcap)) citop citype(logit) format(%9.2f) /// 
 addplot(scatter @b @at, ms(i) mlabel(@b) mlabpos(2) mlabcolor(black))

proportion CasCov_b if dNative==0 & TipoComRev<=2
estimates store nativeb

proportion CasCov_b if dNative==1 & TipoComRev<=2
estimates store campensinob

coefplot nativeb campensinob, vertical recast(bar) barwidth(0.3) fcolor(*.5) ///
 ciopts(recast(rcap)) citop citype(logit) format(%9.2f) /// 
 addplot(scatter @b @at, ms(i) mlabel(@b) mlabpos(2) mlabcolor(black))
graph export coef2.png

[Nick Cox]

I don't know what you want to show, but the mean of a proportion is not the same as either the intercept or the gradient from the kind of regression you showed in #1;