Convert tex numeric variable without value label to string variable

[Chris Jiao]

Hi Statalisters,

I have a basic question. I want to convert variable "srcdate" from numeric to string. However, because this numeric variable is text-form without any value label, so I cannot use -decode- and -tostring- code. Does anyone have a more decent solution?

* Example generated by -dataex-. To install: ssc install dataex
clear
input long srcdate
19358
21549
21184
18382
19358
18627
18992
18474
18839
21549
end
format %d srcdate

Thanks in advanced.

[Chris Jiao]

Hi, when I use -dataex- to demonstrate the data, above, the observations of srcdate are shown in messy code. The correct form that I see in my Stata is the following.

[Maarten Buis]

What you saw is not "messy code", it is the numbers that Stata uses internally (the days since January 1st 1960). The line format %d srcdate controls how those numbers are displayed, in your case it is a date. So 19,358 days after January 1st 1960 is 31st December 2012. Once you have your dates in Stata readable form you almost never want to reverse that. So the big question is: Why do you want to turn this into strings? Maybe we can find an easier, saver way of achieving your final goal.

[Chris Jiao]

What you saw is not "messy code", it is the numbers that Stata uses internally (the days since January 1st 1960). The line format %d srcdate controls how those numbers are displayed, in your case it is a date. So 19,358 days after January 1st 1960 is 31st December 2012. Once you have your dates in Stata readable form you almost never want to reverse that. So the big question is: Why do you want to turn this into strings? Maybe we can find an easier, saver way of achieving your final goal.

Hi, Maarten. I want to merge this dataset with another dataset that contains srcdate variable in string form.

[Chris Jiao]

I tried the following code but it does not work.

gen srcdate_s = date(srcdate, "DMY")

[Chris Jiao]

Hi guys, I made it. Here is the code.

generate srcdate_s = string(srcdate,"%td")

[Maarten Buis]

I think that that is a very dangerous solution, so I would not do that. The problem is that this requires the string in the other dataset always follows the exact same convention. There are so many ways that this can go wrong. "31dec2015" is not the same as "31Dec2015" and not the same as "31 dec 2015" and not the same as "31dec2015 ", etc. etc. etc. The safer solution is to transfer the string in the other dataset to Stata readable format using the date() function. The result will be the "messy code" you saw before. This is a much safer identifier for date, then a a string.