Time in paradata events

Dear all,

I am monitoring a survey and I am interested on check if any interviewer is collecting data after 10pm. Some interviewers have their tablets set up in different time zones. I am interested on converting the time reported in the timestamps to the same time zone.

  1. If an interviewer have a different time zone in her tablet, can I use the variable tz_offset to correct for this issue? For example, Nigerian time is registered as tz_offset==β€œ01:00:00”, but an specific interviewer has tz_offset==β€œ-06:00:00”. Then should I just add 7 hours to the events of this interviewer to harmonize them to the rest of the interviewers?

  2. Should the timestamps in the paradata be the same as the ones in the interview transcript?

For example, see in the images below the time registered for questions s8q1a and s8q1b in the transcript (15:14 -06:00) as well as in the paradata downloaded from the server (21:14:59 with tz_offset -06:00:00)

However, see that for s1q1, the time recorded in the transcript (16:16 -06:00), it is really similar to the one in the paradata (16:14:11 with tz_offset=-06:00:00).

Could you please provide some guidance on how we can harmonize the different time zones?

Thank you so much,


I’m very interested in the answer to this question myself.

For the paradata, I believe that all times are expressed in UTC, and the offset should be applied to that time in order to compute local time (on the device).

Here’s more from the 20.09 release notes:

Paradata export files have the following changes: variable timestamp_utc shows the event timestamp in the UTC using the following format: YYYY-MM-DDThh:mmZ, and variable tz_offset contains the difference in hours between the time zone of the event origin and the UTC. For example, if the timestamp in paradata is recorded as 2020-09-19T16:45Z with time zone offset +3:00 then the clock was showing 19:45 of the same date at the location where the event originated.

To prevent confusion, older variables timestamp and offset are no longer exported. If you have a script or other automated procedure to process paradata, please, adjust accordingly.

For the time in the transcript … :man_shrugging:

Once we figure out what each time represents, I’ll make a PR to the documentation site to explain how to interpret times throughout the system. Future me will thank me :wink:

1 Like

Nigeria is in West African Time Zone:

First and foremost, Any interviewer working in Nigeria must have the tablet set up with this time zone. If the interviewer has an incorrect time zone setting, likely the time settings are also incorrect and any analysis based on that is nonsense. It is the responsibility of the HQ and supervisory staff to make sure every interviewer has the correct clock settings, especially if any analysis will be done based on the time.

In paradata, the timestamps of events are saved into two variables: utc_timestamp and tz_offset, which are the utc time when the event occurred and the time zone offset based on the client’s clock settings.

In transcripts the timestamps are written with the timezone:

For Stata users see also: susotime = code + doc.

No. The time is already harmonized. But harmonized to UTC time. All you have to do is convert your threshold (10pm local time) to UTC, which is going to be 9pm.

Best, Sergiy Radyakin

1 Like

Thank you so much! I am grateful for your help.