Create event time variable

[Ann Jun]

Hi, I have a time series of stock returns with multiple event date indicators, such as the following:

date return event
01jan2015 -0.2 0
02jan2015 1.2 1
03jan2015 0.3 0
04jan2015 0.4 0
05jan2015 -0.3 1
06jan2015 -0.1 0

I would like to create a new variable (timeToEvent) that takes the value of number of days from the event within a 3-day window around each event, where day 0 is each event day. The events are all spread apart by at least 3 days. For instance, it would look like this:

date return event timeToEvent
01jan2015 -0.2 0 -1
02jan2015 1.2 1 0
03jan2015 0.3 0 1
04jan2015 0.4 0 -1
05jan2015 -0.3 1 0
06jan2015 -0.1 0 1

How may I do this?

[Jared Greathouse]

https://www.statalist.org/forums/for...treatment-time

[Hemanshu Kumar]

Here is one solution (which may not be the most efficient) using the community-contributed command rangestat, available from

net install rangestat.pkg

The code:

* ignore the following lines in your own work; they only recreate your sample data

clear
input str9 date float return byte(event timeToEvent)
"01jan2015" -0.2 0 -1
"02jan2015" 1.2 1 0
"03jan2015" 0.3 0 1
"04jan2015" 0.4 0 -1
"05jan2015" -0.3 1 0
"06jan2015" -0.1 0 1
end

* copy the code from here onwards

gen ddate = daily(date,"DMY") // this may not be needed if your original date variable is already in numeric format
gen date_if_event = ddate*event
rangestat (max) eventdate = date_if_event, interval(ddate -1 1)
gen time_to_event = ddate - eventdate
drop ddate date_if_event eventdate

The output is:

. li, noobs ab(20) sep(0)

  +----------------------------------------------------------+
  |      date   return   event   timeToEvent   time_to_event |
  |----------------------------------------------------------|
  | 01jan2015      -.2       0            -1              -1 |
  | 02jan2015      1.2       1             0               0 |
  | 03jan2015       .3       0             1               1 |
  | 04jan2015       .4       0            -1              -1 |
  | 05jan2015      -.3       1             0               0 |
  | 06jan2015      -.1       0             1               1 |
  +----------------------------------------------------------+

You can change the code appropriately if your date variable is not a string. For the future, it is best to provide sample data using the dataex command. That way such confusion about whether some variable is string or numeric, doesn't arise.

[Ann Jun]

Hemanshu Kumar Thank you!