CODEWITHSUNDEEP
stata
Richard Herron
Richard Herron

Reputation: 10112

Determine time period since event (up to n leads/lags)

I want to measure the number of periods (here years) since an event occurred (here represented by indicator variable pos) up to a given number of leads and lags (here three).

The following code works, but seems hackish and like I'm missing something fundamental. Is there a more robust solution that takes advantage of built in functions or a better logic? I'm on 11.2. Thanks!

version 11.2
clear

* generate annual data
set obs 40
generate country = cond(_n <= 20, "USA", "UK")
bysort country: generate year = 1766 + _n
generate pos = 1 if (year == 1776)

* generate years since event (up to three)
encode country, generate(countryn)
xtset countryn year
generate time_to_pos = 0 if (pos == 1)
forvalues i = 1/3 {
    replace time_to_pos = `i' if (l`i'.pos == 1)
    replace time_to_pos = -1 * `i' if (f`i'.pos == 1)
}

Upvotes: 1

Views: 862

Answers (1)

Nick Cox
Nick Cox

Reputation: 37358

Clear question.

This can be shortened. Here is one way. Starting with your code to set up a sandpit 

version 11.2
clear
* generate annual data
set obs 40
generate country = cond(_n <= 20, "USA", "UK")
bysort country: generate year = 1766 + _n

Now it is 

gen time_to_pos = year - 1776 if  abs(1776 - year) <= 3

That is all that seems needed for your example. If you want to generalise to multiple events within each panel, I'd like to know the rules for such events.

I was going to show a trick from http://www.stata-journal.com/article.html?article=dm0055 but it doesn't appear needed.

Upvotes: 2

Related Questions