date variables: some hours written as hh:mm:59 and not rounded

[Michael Duarte Goncalves]

Hi everyone,

I have a very easy question for stata specialists. I have some of my date's hours that are not in the "format" that I want


* Example generated by -dataex-. For more info, type help dataex
clear
input double(date_hour price)
1.9407168e+12 91.99
1940720399999.9998 91.26
1940723999999.9995 91.65
1.9407276e+12 91.99
1940731199999.9998 91.99
end
format %tc date_hour
[/CODE]

For example, the value above in red is written as 01jul2021 00:59:59, instead of 01:00:00.
Basically, my hours should be instead 00:00:00, 01:00:00, 02:00:00, and so on.

Could anyone please give me the solution to that?
Thanks.

Michael

[Nick Cox]

An answer is that date-times are numeric with units ms. Therefore if (and only if) you know all times should be on the hour, push them through round() with second argument 60 * 60 * 1000.

[Michael Duarte Goncalves]

Good afternoon Nick Cox,

Thank you so much for your help, again. In my case, yes, all times should be on the hour.

However, I have a few questions to ask you please:

1. Sorry, what do you mean by second argument?
2. Why are you doing 60*60*1000?

Sorry, I am not proficient yet in datetime variables management.
Thank you in advance for your answer.
Best,

Michael

[Nick Cox]

In a hour there are 60 minutes. In a minute there are 60 seconds. In a second there are 1000 milliseconds.

In a function call such as function(foo, bar) foo is the first argument and bar is the second argument. https://en.wikipedia.org/wiki/Argument_of_a_function

[Michael Duarte Goncalves]

Ok, thank you for your time! All is clear now.

All the best Nick,

Michael

[Michael Duarte Goncalves]

Hi again,

Unfortunately, I still obtain some weird results:

gen test = round(date_hour, 60*60*1000)

dataex:

* Example generated by -dataex-. For more info, type help dataex
clear
input float test
1.9407152e+12
1.9407187e+12
1.9407224e+12
 1.940726e+12
1.9407296e+12
end
format %tc test

Should that be better?

gen double time = round(clock(date_hour, "MDY hms"), 60*60*1000)
format %tc time

Edit: Oh, forget about the code above, date_hour should be a string for the code above to work properly, right?

Thank you again for your help.
Best,

Michael

[Nick Cox]

The first argument of clock() must indeed be a string. Otherwise the idea works for me:

. di %tc  round(1940720399999.9998, 60 * 60 * 1000)
01jul2021 01:00:00

[Clyde Schechter]

I'm not sure what is being said in #6. The key problem is that -gen test = ...- creates test as a float. A float does not have enough bits to store a Stata date time variable without losing precision. It must be -gen double test = ...-. There seems to be some recognition of this in what follows later in that post, but I'm not sure, so I just want to drive that point home. Date time variables in Stata must always be created as -double-s, and doing that requires explicit mention of that in the -gen- statement, because the default is the inadequate (for this purpose) float.

[Michael Duarte Goncalves]

Hi Nick Cox,
Hi Clyde Schechter,

Thank you for your help in #7 and #8.

• #7: It works perfectly as you told me. Thank you!
• #8: You are right Clyde, I forgot to put the -gen double-. So the code written in #6 should be rather:

gen double test = round(date_hour, 60*60*1000)

You said that to me in one of my previous posts. I'm sorry, I completely overlooked this specificity.

Many thanks to both of you for your help.
Have a nice morning/day/evening.

All the best,

Michael