Coefplot: Labelling interactions with a month-year date

[Mariela Dal Borgo]

Hello,

I am using the command coefplot and need to label the axis for an interaction between a time dummy and a continuous regressor. So the model looks like:

reghdfe ddebt i.period#c.shock, a(period state)

I want to label the interaction term i.period#c.shock but only using the date (i.e. the variable "period"), which is formatted as year-month (%tm).

I've read the post: https://www.statalist.org/forums/for...d-interactions

Based on that, I created a regular expression to rename the interaction coefficients as follows:

coefplot, keep(*.period#c.shock) vertical rename(^([0-9][0-9][0-9]).period#c.shock$ = \1, regex)

The problem is I get the labels in numeric format (650, 651), not in date format (2014m3, 2014m4). Is there a way to do it using the option coeflabels()? I couldn't find the way and I have many dates to do it manually.

Thanks for any help!

Mariela

[Andrew Musau]

Can you provide a data example?

[Mariela Dal Borgo]

Hello Andrew,

Yes, I am copying the data and just run the two commands above. Thanks!

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(ddebt period shock) str6 state_name
 0 613  .09739853 "ags" 
 0 612  .09739853 "ags" 
 0 623  .09739853 "ags" 
 0 616  .09739853 "ags" 
 0 621  .09739853 "ags" 
 0 617  .09739853 "ags" 
 0 619  .09739853 "ags" 
 0 614  .09739853 "ags" 
 0 622  .09739853 "ags" 
 0 620  .09739853 "ags" 
 0 618  .09739853 "ags" 
 0 615  .09739853 "ags" 
 0 617  .12131932 "bc"  
 0 612  .12131932 "bc"  
 0 622  .12131932 "bc"  
 0 621  .12131932 "bc"  
 0 616  .12131932 "bc"  
 0 613  .12131932 "bc"  
 0 623  .12131932 "bc"  
 0 618  .12131932 "bc"  
 0 615  .12131932 "bc"  
 0 620  .12131932 "bc"  
 0 619  .12131932 "bc"  
 0 614  .12131932 "bc"  
 0 614  .10386167 "bcs" 
 0 618  .10386167 "bcs" 
 0 615  .10386167 "bcs" 
 0 620  .10386167 "bcs" 
 0 621  .10386167 "bcs" 
 0 613  .10386167 "bcs" 
 0 619  .10386167 "bcs" 
 0 612  .10386167 "bcs" 
 0 622  .10386167 "bcs" 
 0 617  .10386167 "bcs" 
 0 616  .10386167 "bcs" 
 0 623  .10386167 "bcs" 
 1 616  .05919579 "camp"
 1 615  .05919579 "camp"
 1 620  .05919579 "camp"
 1 617  .05919579 "camp"
 1 618  .05919579 "camp"
 1 613  .05919579 "camp"
 1 621  .05919579 "camp"
 1 612  .05919579 "camp"
 1 614  .05919579 "camp"
 1 622  .05919579 "camp"
 1 619  .05919579 "camp"
 1 623  .05919579 "camp"
 0 615   .0766994 "coah"
 0 612   .0766994 "coah"
 0 619   .0766994 "coah"
 0 623   .0766994 "coah"
 0 617   .0766994 "coah"
 0 621   .0766994 "coah"
 0 618   .0766994 "coah"
 0 620   .0766994 "coah"
 0 614   .0766994 "coah"
 0 613   .0766994 "coah"
 0 616   .0766994 "coah"
 0 622   .0766994 "coah"
 0 615  .14029005 "col" 
 0 623  .14029005 "col" 
 0 618  .14029005 "col" 
 0 613  .14029005 "col" 
 0 616  .14029005 "col" 
 0 622  .14029005 "col" 
 0 621  .14029005 "col" 
 0 614  .14029005 "col" 
 0 619  .14029005 "col" 
 0 620  .14029005 "col" 
 0 612  .14029005 "col" 
 0 617  .14029005 "col" 
 0 621  .13756678 "chis"
 0 619  .13756678 "chis"
 0 612  .13756678 "chis"
-1 623  .13756678 "chis"
 0 617  .13756678 "chis"
 0 613  .13756678 "chis"
 0 615  .13756678 "chis"
 0 618  .13756678 "chis"
-1 622  .13756678 "chis"
 0 620  .13756678 "chis"
 0 614  .13756678 "chis"
 0 616  .13756678 "chis"
 0 623 -.02329298 "chih"
 0 613 -.02329298 "chih"
 0 612 -.02329298 "chih"
 0 615 -.02329298 "chih"
 0 617 -.02329298 "chih"
 0 616 -.02329298 "chih"
 0 619 -.02329298 "chih"
 0 622 -.02329298 "chih"
 0 620 -.02329298 "chih"
 0 621 -.02329298 "chih"
 0 618 -.02329298 "chih"
 0 614 -.02329298 "chih"
 0 618 .067410916 "dgo" 
 0 621 .067410916 "dgo" 
 0 623 .067410916 "dgo" 
 0 612 .067410916 "dgo" 
end
format %tm period

[Andrew Musau]

Thanks for the data example. I suppose that having the interactions on the x-axis will not be ideal due to the length of the text. Therefore, I would put the labels on the y-axis. Here, I use labmask from Stata Journal, authored by Nick Cox. reghdfe is from SSC, authored by Sergio Correia and coefplot is also from SSC, authored by Ben Jann.

encode state_name, gen(state)
gen sdate2= string(period, "%tm")
*findit labmask (follow instructions to install)
labmask period, val(sdate2)
reghdfe ddebt i.period#c.shock, a(period state)
coefplot, keep(*.period#c.shock)

Result:

[Mariela Dal Borgo]

Thanks a lot! That was great, the labmask command is very helpful. I also added coeflabel( , truncate(6)) because I don't want the "# shock" in the labels.

[Andrew Musau]

Another way using regular expressions:

 coefplot, keep(*.period#c.shock) rename(#c.shock = \1, regex)

[Mariela Dal Borgo]

It works better, thank you very much!