## data step

Super Contributor
Posts: 323

# data step

In my data set I have two datetime variables in numeric format. I need to calculate difference between the datetime and put the difference into three variables, Forexample,

variable1                   variable2

22oct15:10:47:00       21oct15:10:00:02

the difference between  variable1-variable2=1day 0 hour 47min into three variables  days hour min

1        0      47

I am using the following code.

data one;
set two;

LDDAYS= variable1-variable2;run;

this code is not giving me expected values. Thank you

Super User
Posts: 10,280

## Re: data step

[ Edited ]

datetime values area stored in SAS as numbers, counting seconds from 01jan1960:00:00:00.

Therefore your initial subtraction will give you another datetime value, which will probably be in early January, 1960, if you assign it a datetime format.

I suggest this:

``````diff = variable1 - variable2; * difference in seconds;
lddays = int(diff / 86400); * days;
difftime = diff - lddays * 86400; * subtract the days;
format difftime time8.; * display as time value;``````

---------------------------------------------------------------------------------------------
Maxims of Maximally Efficient SAS Programmers
How to convert datasets to data steps
How to post code
Super Contributor
Posts: 323

## Re: data step

Thank you. That worked. But I need to separate hours and minutes into two variables.

Forexample I got difftime(In redink below code) 1:15:48. I need to place the 1 in hour variable and round of 15.48=16 in Min variable. Thanks for your quick response.

``````diff = variable1 - variable2; * difference in seconds;
lddays = int(diff / 86400); * days;
difftime = diff - lddays * 86400; * subtract thedays;
format difftime time8.; * display as time vale;``````

Super User
Posts: 10,280

## Re: data step

Well, you use the same principle. Use factor 3600 to extract the hours, and 60 to extract the minutes. Use the round() function on the minutes and assign simple numeric formats to the target variables.

---------------------------------------------------------------------------------------------
Maxims of Maximally Efficient SAS Programmers
How to convert datasets to data steps
How to post code
Super User
Posts: 9,599

## Re: data step

I would also suggest, that rather than using a set number, which can be tricky to read if you don't know what that number represents, you can use the SAS functions -> year(), month(), day(), hour(), minute() to extract parts from a date/datetime and use those.

```data want;
variable1="22oct15:10:47:00"dt; variable2="21oct15:10:00:02"dt;
difference_in_days=datepart(variable1)-datepart(variable2);
difference_in_hours=hour(timepart(variable1)-timepart(variable2));
difference_in_minutes=minute(timepart(variable1)-timepart(variable2));
format variable1 variable2 datetime.;
run;```
Super User
Posts: 10,280

## Re: data step

RW9 wrote:
```data want;
variable1="22oct15:10:47:00"dt; variable2="21oct15:10:00:02"dt;
difference_in_days=datepart(variable1)-datepart(variable2);
difference_in_hours=hour(timepart(variable1)-timepart(variable2));
difference_in_minutes=minute(timepart(variable1)-timepart(variable2));
format variable1 variable2 datetime.;
run;```

When doing this, one needs to take care of cases where timepart(variable1)-timepart(variable2) turns out to be negative, and reduce the days by 1.

ie

``````variable1='27jan2016:09:23:15'dt;
variable2='26jan2016:15:45:20'dt;``````
---------------------------------------------------------------------------------------------
Maxims of Maximally Efficient SAS Programmers
How to convert datasets to data steps
How to post code
Super User
Posts: 10,787

## Re: data step

Here is the solution from SAS-L

https://listserv.uga.edu/cgi-bin/wa?A0=SAS-L

``````
proc format;
picture durt
other='%0nD %0H:%0M:%0s' (datatype=time);
run;

proc fcmp outlib=work.func.fmt;
function xx(c) \$;
length fmt \$ 20;
day=int(divide(c,60*60*24));
return (put(day,z2.) || 'D ' || put(c,tod.));
endsub;
run;
options cmplib=(work.func);
proc format;
value new(default=20) other=[xx()];
run;
data _null_;

t= 368000;
put t durt. / t new.;
run;
``````
```OUTPUT:
4D 06:13:20
04D 06:13:20 ```
Discussion stats
• 6 replies
• 331 views
• 0 likes
• 4 in conversation