sort dataset on date variable

[Ingebrigt Meisingset]

Hi, I want to sort a dataset according to the date for the observations. I have a variable date in this format: "DMYhms". One examle of the variable: "05-01-2017 08:54:45". The command sort uses only the first part of the variable (Day) to sort the dataset. How can I sort the dataset from the newest to the oldest obserervation?

[Nick Cox]

I think what you say is wrong as sort will not ignore any part of the string it is sorting on.

But converting this to a Stata date-time variable is long overdue.

gen double numdate = clock(strdate, "DMY hms")
gsort -numdate 

[Ingebrigt Meisingset]

Thanks!

[Martin Imelda Borg]

Hi Nick Cox
Assuming one has converted the stata date into a Stata form using the code:

gen admissiondate = date(admidate, "YMDhms")

will the sort only work if the date it is in STATA form?

I find that once I have converted the dates into a STATA date and then into a readable form for the human eyes, stata will sort the dates but it's not the ascending order...
format admissiondate %td

[Nick Cox]

#4 Please give a data-based example, FAQ Advice #12 applies (see also #18 while visiting).

[Rich Goldstein]

the code in #4 appears wrong in two senses, each covered in @Nick Cox's post in #2: (1) your generate statement should be followed by the word "double" as highlighted in the help file; (2) if you want reversed order you should, as Nick did in #2, use gsort with a minus sign before the name of the variable you wanted reverse sorted; again, see the help file for -gsort-

[Jared Greathouse]

You can just do

sort -date

If I understand you well

[Nick Cox]

Too much small confusion here. Working backwards

#7 Jared For that you need gsort not sort.

#6 Rich Indeed. Also there may be confusion in #4 between clock() and date() -- noting that date() yields daily dates only, and not date-times.

[Martin Imelda Borg]

Too quick !

Aim: To sort generate a _n observation for each episode the patient was discharged in hospital and sort the episodes according to discharge date in ascending order

Here's a copy of my data:
----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. For more info, type help dataex
clear
input float(Procedureno Indexno HospitalD discharge episodeno)
20 1 17 14762 1
 . 0 17 14396 2
21 1 20 14335 1
21 1 20 22743 2
end
format %td discharge

------------------ copy up to and including the previous line ------------------

Listed 4 out of 4 observations

Code used:
bys HospitalD (Procedureno Indexno discharge): gen episodeno = _n

The final result of the code is above.
I thought this was secondary to the fact that perhaps I translated into into readable human form and therefore stata couldnt work it out
I tried the above again just with stata dates (not translating to human form) and it still didn't sort it out...

[Martin Imelda Borg]

Perhaps I should post this on a separate thread ?
but As we’re talking about sorting dates I thought this would be relevant here