Multiple versions - Randomization?

I have 20 different versions of the first question question. Remaining part of the study remains the same. A subject has to be given only 1 version of the first question and the remaining questions are the same for all participants.

1 option I have is to create 20 different input files (one with a different version for Q1) and append the remaining common questions.

Other option is having a single input file and to find a way to randomize the 20 versions within group (displaying only one of them) and then displaying the common questions.

How can I accomplish this? Any thoughts/ ideas will be appreciated.

Thank you.

Monil C

I would go with the latter (assuming you mean with replacement or "true" randomization). You can add an unconditional skip to each of the first twenty (randomized) items such that no matter which comes first, the participant is skipped past the other 19.

Thank you Jarvis I will try that.

I am wondering if there is any statistics about the randomness provided by the software? I mean, I am curious to see how will the study perform in terms of selecting each of the 20 different versions in the most uniform manner.

Is there some test input file you have which will help me record the randomness or something?

Thank you.

Best Regards,
Monil Chheda

Although I have done distribution tests in the past, I don't have anything like that in a postable form. The randomization uses VB6's randomization function as well as the "randomize" function to avoid repetitions in random orders between sessions. FWIW, I've always been satisfied with the random distributions. If you like, Google "vb6 randomize" and I'm sure you will find plenty of discussion.

Note also that you could start off with a question like "what is the condition?" and then skip to a specific trial (from the 20) based on that and then skip from there to the rest of the study. That would allow you to keep n sizes equal.