Coefplot multiple variables in mutliple models

[rouault eleonor]

Dear Stata users,

I want to coefplot two variables (var1_x and var2_x) in the same graph across different models. For each month of the year (noted x), I run one regression. Thus, I run the following set of regressions:

reg y var1_1 var2_1 controls
eststo Jan
reg y var1_2 var2_2 controls
eststo Feb
reg y var1_3 var2_3 controls
eststo Mar
...
reg y var1_12 var2_12 controls
eststo Dec

coefplot  (jan) (feb) (ma) ... (dec), keep(var1* var2*) ylabel aseq swapnames xline(0)

The problem is that I obtain two labels for each monthly variable and a doubled y-axis scale in the coefplot. How can I have only one label per variable in a unique coefplot (and only one scale for the whole graph)?

Sorry if the topic has already been treated; I don't find any help in the forum.

[Andrew Musau]

coefplot is from SSC (FAQ Advice #12). Reproducible example??

[Scott Merryman]

First, -ylabel- is not an option in coefplot and isn't this results of the -swapname- option?

Compare

clear*
webuse grunfeld
qui {
    reg invest mvalue kstock if year == 1935
    eststo y1
    reg invest mvalue kstock if year == 1936
    eststo y2
    reg invest mvalue kstock if year == 1937
    eststo y3
    reg invest mvalue kstock if year == 1938
    eststo y4
    reg invest mvalue kstock if year == 1939 
    eststo y5
}
coefplot  (y1) (y2) (y3) (y4) (y5),  /// 
    keep(mvalue kstock) aseq xline(0) name(gr1,replace) 

coefplot  (y1) (y2) (y3) (y4) (y5),  /// 
    keep(mvalue kstock)  swapnames xaseq line(0)

[rouault eleonor]

Thank you for your responses.

Unfortunately, the option xaseq seems to be not allowed (r(198) error).

This is more precisely what I want:




I tried to reshape the data you use Scott, to adapt to what I look for.

clear*
webuse grunfeld
keep if company==1
expand 12
bysort year : gen month = _n
qui {
    reg invest mvalue kstock if month== 1
    eststo jan
    reg invest mvalue kstock if month== 2
    eststo feb
    reg invest mvalue kstock if month== 3
    eststo mar
    reg invest mvalue kstock if month== 4
    eststo apr
    reg invest mvalue kstock if month== 5
    eststo may
}
coefplot  (jan) (feb) (feb) (mar) (apr) (may),  ///
    keep(mvalue kstock) aseq name(gr1,replace)

But mvalue and stock are separated. I would prefer for each month to have both variables estimates.

Maybe I should rely on matrix ?

I'm using Stata16