Extracting quarter from year quarter date

[Guest]

Dear All,

I got stuck with my work. I have year-quarterly variable t e.g. 2020q1. I wanted to extract only quarter from it. I used the command:

gen quarter=qofd(yofd(t))

Stata shows me that everything is 21 quarter...
I don't understand the logic behind this. I wanted to extract year and used the code:

gen year=yofd(dofq(t))

and got the result.

[Nick Cox]

quarter() feeds on daily dates, so you need to map quarterly dates to daily dates first.

. di yq(2022, 2)
249

. di %tq  yq(2022, 2)
2022q2

. di quarter(dofq(yq(2022, 2)))
2

Don't shoot the messenger. Alternatively extrdate from numdate from SSC does it all for you.

. gen have = yq(2022, 2)

. format have %tq


. extrdate quarter  quarter = have


. l have quarter in 1

     +------------------+
     |   have   quarter |
     |------------------|
  1. | 2022q2         2 |
     +------------------+

[Andrew Musau]

You can exploit the fact that the quarter is the last digit when displaying a quarterly date variable.

* Example generated by -dataex-. For more info, type help dataex
clear
input float quarter
198
244
245
end
format %tq quarter

gen wanted= real(substr(string(quarter,"%tq"), -1, 1))

Res.:

. gen wanted= real(substr(string(quarter,"%tq"), -1, 1))

. l

     +------------------+
     | quarter   wanted |
     |------------------|
  1. |  2009q3        3 |
  2. |  2021q1        1 |
  3. |  2021q2        2 |
     +------------------+

Note: Crossed with #2.

[Nick Cox]

Alternatively the quarter is 1 + mod(quarterly_date, 4).

[Guest]

Thank you Nick and Andrew for your answers. When I typed gen quarter=quarter(dofq(yq(time))) I saw invalid syntax.
Unfortunately I cannot install numdate. When I typed ssc install numdate I saw: host not found.
I tried to understand how the command in the #4 works. I found that mod(x,y) produces remainders of x from division by y.
Since a year has 4 quarters I understand we have to divide time by 4. I don't understand why we are adding 1.

Anyway, I got my result and I am very grateful for your kind help!

[Nick Cox]

yq() requires two arguments -- the year and the quarter -- and you supplied one.

If you divide integers by 4 then there are 4 possible remainders, namely 0 1 2 3. That is why you need to add 1 because you want to map to quarters 1 2 3 4.

Here is that done more slowly. I created a sandbox dataset in which we know the quarter. But the point is how do we extract the quarter from a quarterly date?

. clear

. set obs 8
Number of observations (_N) was 0, now 8.

. gen year = cond(_n <= 4, 2020, 2021)

. egen quarter = seq(), to(4)

. gen qdate = yq(year, quarter)

. format qdate %tq

. list, sepby(year)

     +-------------------------+
     | year   quarter    qdate |
     |-------------------------|
  1. | 2020         1   2020q1 |
  2. | 2020         2   2020q2 |
  3. | 2020         3   2020q3 |
  4. | 2020         4   2020q4 |
     |-------------------------|
  5. | 2021         1   2021q1 |
  6. | 2021         2   2021q2 |
  7. | 2021         3   2021q3 |
  8. | 2021         4   2021q4 |
     +-------------------------+

. gen remainder = mod(qdate, 4)

. gen wanted = remainder + 1

. list, sepby(year)

     +---------------------------------------------+
     | year   quarter    qdate   remain~r   wanted |
     |---------------------------------------------|
  1. | 2020         1   2020q1          0        1 |
  2. | 2020         2   2020q2          1        2 |
  3. | 2020         3   2020q3          2        3 |
  4. | 2020         4   2020q4          3        4 |
     |---------------------------------------------|
  5. | 2021         1   2021q1          0        1 |
  6. | 2021         2   2021q2          1        2 |
  7. | 2021         3   2021q3          2        3 |
  8. | 2021         4   2021q4          3        4 |
     +---------------------------------------------+

[Guest]

Thank you for such a thorough explanation! Now I got it