Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Randomization while fielding?

Hey guys-

I have a client who wants to add in block randomization to a survey that is currently live and collecting data.

If I used the block randomization feature, would there be any implications from doing so? (AKA- can I just add it and upload the study with no consequences?)
asked Mar 28 by anonymous

1 Answer

0 votes
A few issues to note and address up front ...

1/ Backup your data just before making any changes. Best to be safe hey!

2/ Your current list of completed and disqualified surveys are safe. No issues with them at all.

3/ If you make the changes, new surveys that have not started yet will see the new survey version (with the randomised blocks implemented). These will be fine also.

4/ The concerns surround the suspended / incompleted surveys. I would take a snapshot of the incompletes within the admin module just before you make the survey changes. You also have a backup of the data too. Consider the incompleted surveys that are suspended around the questions you are applying block randomisation to. These are the likely surveys that may encounter some missing data. Take a note of their sys_RespNum and monitor their activity / data through to the end of field work. Incompleted surveys that have not reached the block randomisation questions, or have passed them, will not encounter any issues.

5/ To summarise, I would make the changes at a quite surveying period. Around this time, backup your data, take a snapshot of the incompletes and note any sys_RespNum that hover around the block randomisation questions.

6/ Your new survey version data will have system generated data associated to the block randomisation that the version 1 survey will not have.

Does that all make sense? Let me know if you seek further clarification on any of the above.

Regards Paul. Stay safe and well.
answered Mar 28 by Paul Moon Platinum (80,455 points)
Makes sense-

But to this point-
"Your new survey version data will have system generated data associated to the block randomisation that the version 1 survey will not have."
That's not referring to actual variables in the survey? So, my output will look the same, except for something outside of my variables being  attached to the end of the data set?
Yes, correct. Your survey data will all be the same but some new block randomisation data will appear in your data set for the version 2 surveys only. This is is fine.
You will see a system field like this "sys_block_set_1" appear in your new version 2 data and if you use a constructed list to assist with the block randomisation, the data for that constructed list will also be included.

So this is additional system data for the version 2 survey. Your actual survey questions data will be the same for versions 1+2.
And if I have two blocks within my survey that I want to randomly assign such that each respondent only sees one of them- I use a constructed list? Still no implications for data?
That will be fine. Your constructed list will have some SSI script or Perl script that will select only block 1 or block 2.