How to draw two base lines using coefplot command on stata.

[Aiman Farrukh]

Hi,

I'm trying to look at the average effect in the pre- and post treatment time periods. For that I wish to plot 2 base lines (average of coefficients obtained), one for the pre-treatent time periods and other for the post treatment time periods using the coefplot command on stata. Could someone help me understand how do I get two baselines in the same graph using xline condition or something of that sort?

[Scott Merryman]

I am not too familiar with -coefplot-, but you can have multiple values in the -xline()- option

sysuse auto
regress price mpg
coefplot, xline(-225 -275) drop(_cons)

[Aiman Farrukh]

Thank you for replying Scott. But I did not correctly write my query. I want to create two horizontal reference lines similar to the picture attached with this post. The reference lines cut in between. If you've any idea then please help.

[Scott Merryman]

There might be a way to doing this in -coefplot- but here in another way using Roger Newson's -parmby- to collect the estimation results (-ssc desc parmest- for the various modules make up the parmest package).

webuse grunfeld,clear
qui parmby "reg invest mvalue" ,by(year) fast

twoway connect est year if parm == "mvalue" , mc(black) lp(dash)  /// 
|| rcap max min year if parm == "mvalue", lc(gs10) /// 
|| scatteri .09 1935 .09 1945, recast(line) lw(thick) lp(dash)  lc(gs6) /// 
|| scatteri .19 1946 .19 1955 , recast(line) lw(thick) lp(dash) lc(gs6) /// 
|| , legend(off) xline(1945.5, lp(solid) lw(medthick) lc(black))

[Aiman Farrukh]

Thank you so much for your response. I have managed to create the reference lines.
However, I'm unable to plot the coefficients. Attached with this post a picture of name of all the coefficients by year.

[Aiman Farrukh]

This is the command I've used:

qui parmby "reghdfe work 1.treat#ib(1999).year zero if rural==1 & sex==0 & age>14 & age<60, absorb(district year age_cat)vce(cluster district)", by(year) fast

[Scott Merryman]

You might want to think about your specification. You have year dummy variables interacted with treatment variable, year fixed effects, and you are going each regression by year.