Dummy variables for day of the week and month of the year

[Raul Athwall]

Hi, I am trying to deseasonalise my data and want to create dummy variables for the days of the week and all the months. My data is numerical so how do I create weekday variables when I do not know which data corresponds to which day?

[Nick Cox]

Assuming that you have daily dates in a variable whose name you don't give, this script shows you some technique:

. clear 

. set obs 365 
Number of observations (_N) was 0, now 365.

. gen ddate = mdy(12,31,2021) + _n

. 
. gen dow = dow(ddate)

. gen month = month(ddate)

. 
. tab month, gen(month) 

      month |      Freq.     Percent        Cum.
------------+-----------------------------------
          1 |         31        8.49        8.49
          2 |         28        7.67       16.16
          3 |         31        8.49       24.66
          4 |         30        8.22       32.88
          5 |         31        8.49       41.37
          6 |         30        8.22       49.59
          7 |         31        8.49       58.08
          8 |         31        8.49       66.58
          9 |         30        8.22       74.79
         10 |         31        8.49       83.29
         11 |         30        8.22       91.51
         12 |         31        8.49      100.00
------------+-----------------------------------
      Total |        365      100.00

. 
. d 

Contains data
 Observations:           365                  
    Variables:            15                  
-----------------------------------------------------------------------------------------
Variable      Storage   Display    Value
    name         type    format    label      Variable label
-----------------------------------------------------------------------------------------
ddate           float   %9.0g                 
dow             float   %9.0g                 
month           float   %9.0g                 
month1          byte    %8.0g                 month== 1.0000
month2          byte    %8.0g                 month== 2.0000
month3          byte    %8.0g                 month== 3.0000
month4          byte    %8.0g                 month== 4.0000
month5          byte    %8.0g                 month== 5.0000
month6          byte    %8.0g                 month== 6.0000
month7          byte    %8.0g                 month== 7.0000
month8          byte    %8.0g                 month== 8.0000
month9          byte    %8.0g                 month== 9.0000
month10         byte    %8.0g                 month== 10.0000
month11         byte    %8.0g                 month== 11.0000
month12         byte    %8.0g                 month== 12.0000
-----------------------------------------------------------------------------------------

For most modelling purposes you don't need indicator variables but can just use factor variable notation, so new variables like dow and month are the only crucial steps.

At some point, you can waste more time than you save by not reading the documentation, here

help datetime 

to explain basics about dates of various kinds. It is a good idea still to skim and skip what you don't need right now, nevertheless noting that there are details you can come back to.

[Raul Athwall]

Thank you so much Nick!

[Raul Athwall]

Just a question, I have over 1 million observations as this is time series data- setting the obs as 365 does not work for some reason, is there a way to fix this?

[Nick Cox]

#3 Glad it helped.

#4 Correct; that does not work for you.

You didn't give an example in #1 (compare our request at https://www.statalist.org/forums/help#stata) -- so I created one myself with 365 observations, although the point could have been made differently too.

But set obs 365 is both unnecessary and wrong in your case:

1. unnecessary, because you already have data in memory and don't need to create a data example.

2. wrong, because Stata will only allow you to use set obs to make a dataset bigger (including the very useful case of bigger than 0 observations) and set obs 365 could only imply a contraction of your dataset.

Typing

help set obs

would have led you to an explanation. Please see also https://www.statalist.org/forums/help#before