Page 1 of 1

comparing "random allocation of tasks" and "block"

PostPosted: Thu May 19, 2016 12:51 pm
by Yana
Hi,
This question is related with web survey. Thinking about two strategies in letting respondent only seeing part of the design so that they do not feel fatigue: block, and random allocation.
Block means to split the design into several versions of questionnaire. Random allocation is a different but also feasible function( and sometime less costly) when doing survey on web: let's say the design has 24 tasks, and we randomly allocate each respondent with 8 out of the 24 tasks.
Can I say random allocation is even better because it would be less restrictive than using the blocking?
Or random allocation is likely to have some problems? Because I read In chapter 6.2.3.6 of the book of Applied Choice Analysis it says: "they analyst may elect to randomly allocate treatment combinations to different decision makers. While this strategy may result in a confoundment between the treatment combinations given to decision makers and the parameter estimates......"
I don't understand why this may result in confoundment.

Many thanks!

Re: comparing "random allocation of tasks" and "block"

PostPosted: Fri May 20, 2016 9:40 am
by Michiel Bliemer
Blocking is typically considered better since it makes sure that respondents see an equal number of different levels over all choice tasks. So within each block the attribute levels are balanced (as much as possible anyway) such that respondents do not see just high levels or low levels. If you allocate randomly, it may be that certain respondents to not see certain levels or certain combinations of levels, which may affect their choices.

So I would say preferably use blocking over random allocation, but I think in many cases using random allocation will work fine without many problems, so I would not worry too much about it.