Subtracting dates to get day intervals

[Json Ochs]

Hi all, thanks again for taking the time to answer my beginner questions

I am trying to determine the difference between hospital discharge and the date of tests for a large dataset.

date of discharge variable = "date" and is in string format (str10)
date of test variable = "stress_date" and is in string format (str10)
both are in MDY format (e.g. 10/28/2021)

I have tried the below which generated 1189 missing variables
. generate time_to_stress = date("stress_date","MDY") - date("date","MDY")

but then tabulating the new variable yields no observations
. tab time_to_stress

I cant seem to wrap my head around what Im doing wrong. When I try the codebook function on "stress_date" I'm not given examples, so maybe the problem is there. I appreciate your time and advice. Thanks


Log:

. generate time_to_stress = date("stress_date","MDY") - date("date","MDY")
(1,189 missing values generated)

. tab time_to_stress
no observations

. codebook date

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------
date encounter date
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Type: String (str10)

Unique values: 197 Missing "": 0/1,189

Examples: "11/14/2021"
"12/21/2021"
"7/29/2021"
"8/4/2021"

. codebook stress_date

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------
stress_date Stress Test Date
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Type: String (str10)

Unique values: 31 Missing "": 1,157/1,189

Examples: ""
""
""
""

. tab stress_date

Stress Test |
Date | Freq. Percent Cum.
------------+-----------------------------------
03/02/2022 | 1 3.13 3.13
03/08/2022 | 1 3.13 6.25
1/11/22 | 1 3.13 9.38
10/1/2021 | 1 3.13 12.50
10/1/21 | 1 3.13 15.63
10/13/21 | 1 3.13 18.75
10/19/21 | 1 3.13 21.88
10/20/2021 | 2 6.25 28.13
10/22/2021 | 1 3.13 31.25
10/22/21 | 1 3.13 34.38
10/25/2021 | 1 3.13 37.50
10/27/2021 | 1 3.13 40.63
10/27/21 | 1 3.13 43.75
10/29/2021 | 1 3.13 46.88
11/17/21 | 1 3.13 50.00
11/2/21 | 1 3.13 53.13
12/2/21 | 1 3.13 56.25
2/12/2022 | 1 3.13 59.38
2/14/22 | 1 3.13 62.50
3/2/22 | 1 3.13 65.63
4/1/22 | 1 3.13 68.75
7/30/21 | 1 3.13 71.88
8/2021 | 1 3.13 75.00
8/25/2021 | 1 3.13 78.13
8/27/21 | 1 3.13 81.25
8/30/21 | 1 3.13 84.38
9/15/21 | 1 3.13 87.50
9/27/21 | 1 3.13 90.63
9/28/2021 | 1 3.13 93.75
9/28/21 | 1 3.13 96.88
9/8/21 | 1 3.13 100.00
------------+-----------------------------------
Total | 32 100.00

[Nick Cox]

You need e.g.

date(stress_date,"MDY")

not

date("stress_date","MDY"

Giving the string "stress_date" as the first argument to date() is legal but not what you you want. The literal string "stress_date" doesn't contain any date information. You need to pass the variable name.

To see this if it's a little subtle still: You could try some arbitrary text say "cats and dogs" that contains no date information and you will get missing too.

. di date("cats and dogs", "MDY")
.

In short, Stata is telling you that no date is defined by what you supplied, but your code was not in error about syntax.

[Json Ochs]

thank you very much