Date format with decorrelation

[Oyvind Roth]

Hey!

I currently have a dataset where I am trying to decorrelete 5 price variables on spot price for electricity on 2 variables of fixed price electricity. I am quite new to Stata, but I have two variables where the first variable A shows the year, for example 2023, and the second variable B shows the year and week, for example 2023-1. What I have been trying is to split the B variable up into B1 and B2 and then add them together to get a date variable to use like this:

split B, parse ("-")
gen date = date(B1 + "w" + B2, "Yw")

But when I do this I am only getting dots in the date variable when browsing through the data (missing values)? And when I see over B1 and B2 they look fine. After that my idea was to use pca after I could get tsset date to work. Do anyone see the problem and could point me in the right direction?

[Andrew Musau]

* Example generated by -dataex-. To install: ssc install dataex
clear
input str6 B
"2020-1"
"2020-2"
"2020-3"
"2020-4"
"2021-1"
end

gen wanted= weekly(B, "YW")
format wanted %tw

Res.:

. l

     +-----------------+
     |      B   wanted |
     |-----------------|
  1. | 2020-1   2020w1 |
  2. | 2020-2   2020w2 |
  3. | 2020-3   2020w3 |
  4. | 2020-4   2020w4 |
  5. | 2021-1   2021w1 |
     +-----------------+

[Nick Cox]

Note that weekly() knows nothing about any definition of week which implies that sometimes there are 53 weeks in a year. If that is a problem you need some other way to deal with them.

Other way round date() is, and is only, a way to extract daily dates from string input, but that is not what you want.

[Nick Cox]

A Stata user and two R users walked into a bar.

[Nick Cox]

Good jokes always seem to involve three people.

(Sorry about the triple posting.)

[Eric Makela]

Perhaps this improvement works?

split B, parse("-") destring
gen timevar=yw(B1,B2)