Coefplot different colours confidence intervals

[Joan Marti]

Hi, I am creating a coefficient plot with confidence intervals at the 99, 95, 90 percent significance intervals. I would like the confident intervals to be different colours depending on the significance level

sysuse auto
eststo reg_1: reg price mpg
eststo reg_2: reg price mpg headroom
coefplot reg_1 reg_2, keep(mpg) levels(99 95 90)

I tried adding the following:

coefplot (reg_1, mcolor(ebblue) ciopts(color(ebblue))) (reg_2, mcolor(ebblue) ciopts(color(ebblue))), keep(mpg) levels(99 95 90)

but it seems only the 99% confidence interval is affected, how can I chose the colour of each confidence interval?

[Nick Cox]

Cross-posted at https://stackoverflow.com/questions/...vals-by-colour

coefplot is a great command I never use -- because I just don't write papers using that kind of plot. More importantly, it is a very popular command but none of its users answered this, so the best advice is to email its author Ben Jann directly. But I can't speak for Ben's policy on user support. He has written so many excellent and popular commands that he may not be able to provide one-to-one support.

[Chen Samulsion]

A chunk of repeated codes and not sure whether it is what you need:

coefplot (reg_1, mcolor(ebblue) ciopts(color(ebblue)) level(90)) (reg_1, mcolor(navy) ciopts(color(navy)) level(95)) (reg_1, mcolor(maroon) ciopts(color(maroon)) level(99)) (reg_2, mcolor(ebblue) ciopts(color(ebblue)) level(90)) (reg_2, mcolor(navy) ciopts(color(navy)) level(95)) (reg_2, mcolor(maroon) ciopts(color(maroon)) level(99)), keep(mpg) legend(row(2))

[Joan Marti]

Thank you Chen, however I was trying to have the three confidence intervals on top of each other, like in the example code I sent, do you think it is possible?

[Chen Samulsion]

Maybe you can try this (I use more distinguishable color instead of navy and maroon in #3):

coefplot (reg_1, mcolor(ebblue) ciopts(color(ebblue yellow red)) level(99 95 90)) ///
         (reg_2, mcolor(ebblue) ciopts(color(ebblue yellow red)) level(99 95 90)), keep(mpg)

see #6

[Chen Samulsion]

Maybe you can try this (I use more distinguishable color instead of navy and maroon in #3):

coefplot (reg_1, mcolor(ebblue) ciopts(color(ebblue yellow red)) level(99 95 90)) (reg_2, mcolor(ebblue) ciopts(color(ebblue yellow red)) level(99 95 90)), keep(mpg)

[Joan Marti]

this is perfect! thank you very much!

[Joan Marti]

Also, in this case, if I distinguish both regression models by changing the shape of the coefficient interval (e.g. msymbol(t) in reg_1 and msymbol(s) in reg_2), is there a way to have two legends, one showing the distinction by colour and other other one by msymbol? All I have managed to do is ammend the existing legend which is based on the regression models but ideally I guess the best approach would be to create a custom legend from scratch (if at all possible)?


thank you very much again