Changing a time variable defined as a string

[Aziz Essouaied]

Dear Stata users, hello;

I'm working with this data currently:

* Example generated by -dataex-. To install: ssc install dataex
clear
input str7 periode float dettepubliqueendupib
"2000-T1"  60.5
"2000-T2"  60.4
"2000-T3"  59.8
"2000-T4"  59.7
"2001-T1"  58.7
"2001-T2"  59.6
"2001-T3"  59.7
"2001-T4"  59.3
"2002-T1"  59.6
"2002-T2"  60.9
"2002-T3"  60.7
"2002-T4"  61.3
"2003-T1"  62.7
"2003-T2"  64.9
"2003-T3"  65.4
"2003-T4"  65.4
"2004-T1"  65.9
"2004-T2"  67.3
"2004-T3"    67
"2004-T4"  66.9
"2005-T1"  67.8
"2005-T2"  68.2
"2005-T3"  67.9
"2005-T4"  68.2
"2006-T1"  68.1
"2006-T2"    67
"2006-T3"  66.9
"2006-T4"  65.4
"2007-T1"  66.5
"2007-T2"  67.4
"2007-T3"  66.8
"2007-T4"  65.5
"2008-T1"  66.8
"2008-T2"  67.3
"2008-T3"  67.8
"2008-T4"  69.8
"2009-T1"  74.7
"2009-T2"  79.8
"2009-T3"  82.5
"2009-T4"  84.1
"2010-T1"  85.9
"2010-T2"  88.3
"2010-T3"  86.4
"2010-T4"  86.3
"2011-T1"  87.7
"2011-T2"  89.2
"2011-T3"  88.4
"2011-T4"  88.7
"2012-T1"  91.2
"2012-T2"  92.6
"2012-T3"  91.3
"2012-T4"  91.7
"2013-T1"  93.3
"2013-T2"  94.8
"2013-T3"    94
"2013-T4"  94.5
"2014-T1"  96.1
"2014-T2"  96.7
"2014-T3"  96.5
"2014-T4"  96.1
"2015-T1"    98
"2015-T2"  98.2
"2015-T3"  97.7
"2015-T4"  96.9
"2016-T1"  98.8
"2016-T2"  99.9
"2016-T3"  99.1
"2016-T4"  98.1
"2017-T1" 100.8
"2017-T2" 101.2
"2017-T3" 100.2
"2017-T4"  98.7
"2018-T1"  99.1
"2018-T2"  99.1
"2018-T3"  99.3
"2018-T4"  98.5
"2019-T1"  99.3
"2019-T2"    99
"2019-T3"  99.9
"2019-T4"  98.1
"2020-T1" 101.2
"2020-T2"   113
"2020-T3" 115.2
"2020-T4" 114.8
"2021-T1" 117.7
"2021-T2" 114.5
"2021-T3" 115.5
"2021-T4" 112.7
"2022-T1" 114.2
"2022-T2" 112.8
"2022-T3"   113
"2022-T4" 111.2
"2023-T1" 111.9
"2023-T2" 111.2
"2023-T3"   111
"2023-T4" 109.9
"2024-T1" 110.5
"2024-T2"   112
end

The data describes the evolution of publc debt of a country from the first year quarter of 2000 to the second year quarter of 2024. The first variable is the time variable defined as "year-quarter", and the second variable is the value of the public debt. As you can see, the "periode" time variable is defined as a string, but I want to change it in a way that I could apply the tsset command on it and treat it as a time variable defined as quarters of the years.

Any help please?

With many thanks!

[Andrew Musau]

gen qdate= quarterly(periode, "YQ")
format qdate %tq
tsset qdate

Res.:

. l in 1/10

     +-----------------------------+
     | periode   dettep~b    qdate |
     |-----------------------------|
  1. | 2000-T1       60.5   2000q1 |
  2. | 2000-T2       60.4   2000q2 |
  3. | 2000-T3       59.8   2000q3 |
  4. | 2000-T4       59.7   2000q4 |
  5. | 2001-T1       58.7   2001q1 |
     |-----------------------------|
  6. | 2001-T2       59.6   2001q2 |
  7. | 2001-T3       59.7   2001q3 |
  8. | 2001-T4       59.3   2001q4 |
  9. | 2002-T1       59.6   2002q1 |
 10. | 2002-T2       60.9   2002q2 |
     +-----------------------------+

[Aziz Essouaied]

Andrew Musau Thanks for the help! It worked.

[Nick Cox]

In short, quarterly() works fine even though your string variable does not include q or Q. But why does it work?

One optimistic idea is that the function is smart enough to recognise T (presumably meaning here "trimestre") as an acceptable abbreviation (in French) indicating quarter.

In English too, "trimester" has the meaning of a three-month period, but to my understanding as referring most often to the first, second, and third three months of a pregnancy lasting nine months or so. So, that meaning has nothing to do with quarters of a calendar year.

Some experiments indicate otherwise. quarterly() is just ignoring non-numeric text, not being smart about what that text is. For example, quarterly("2000-M5", "YQ") would yield missing as quarter 5 isn't defined. But quarterly("2000-M4", "YQ") would yield the same result as quarterly("2000-Q4", "YQ") or quarterly("2000-T4", "YQ").

[Aziz Essouaied]

Nick Cox Thanks for the response. Yes, I do mean quarters of the year, the group of three months. You are right, the word "trimester" is not that very much used, even more in a professional or academic context. But here, in the case of my data, it has nothing to do with pregnancy.