Creating a loop to count the weeks before and after a policy

[Lindsey Yates]

Hi,

I'm working on understanding the impact of a hospital policy on healthcare utilization. The policy was implemented in March of 2020. I need to create variable that increases by 1 for every week after the policy and decreases by 1 for every week before the policy. I know I could do it by creating a lot of replace statements e.g.

gen policy_week=0 if week==11 & year==2020
replace policy_week=1 if week==12 & year==2020
replace policy_week=2 if week==13 & year==2020
replace policy_week=3 if week==14 & year==2020
replace policy_week=4 if week==15 & year==2020
.
.
.
replace policy_week=-1 if week==10 & year==2020
replace policy_week=-2 if week==9 & year==2020
replace policy_week=-3if week==8 & year==2020
replace policy_week=-4 if week==7 & year==2020

But I would like to try it doing a loop. Any guidance on how to best do this?

[Nick Cox]

For your examples, this alone, and no loop, is needed

gen policy_week = week - 11 if year == 2020

but a deeper problem is how many years you have and how you are holding what appear to be weekly data.

Stata's weekly dates aren't obviously the solution without more detail on what weeks mean to you, together with years. For example, if there can be a week 53 and/or if weeks can span two years, then you need something else.

[Lindsey Yates]

Thank you for the clarification, Nick. Yes, the time spans two years from March 2019 - March 2021. The policy was implemented on March 15, 2021, but because I don't have data for everyday I was thinking I could roll up the daily dates to week. I was just reading up on your 'Stata tip 68: Week assumptions' to better understand some of the limitations weeks in stata. What is the better solution?

[Nick Cox]

In general by far the best strategy is to use daily dates if you have them, so your counter is just a daily date MINUS mdy(3, 15, 2021)

[Lindsey Yates]

Thank you Nick! I meant that the policy was implemented on 3/15/2020 (not 2021), but the code worked.