choice-metrics.com

Hi there,

Just wondering if it is theoretically okay to block an efficent design; I have scribbled down from the course that blocking is technically only relevant for orthogonal designs. However for cost and practical reasons I would like to reduce the number of choice sets seen and am reluctant to reduce attribute numbers, levels etc any further. Any advice offerred would be much appreciated

Cheers

Tracey

Hi Tracey,

It is ok to block an efficient design. Originally, blocking was 'invented' for orthogonal designs, as they tend to be very large and blocking is necessary. Although efficient designs are usually much smaller, they can also still be too large to completely give to a single respondent. Blocking ensures in an orthogonal design that one respondent will observe all levels an equal number of times, hence not only low levels or high levels of a certain attribute. This is done by finding another orthogonal column. Since an efficient design in general is not orthogonal, it is not possible to find such a blocking column that is completely orthogonal to all other columns, but Ngene finds one that minimizes the correlations to all other columns, i.e. near-orthogonal. In that sense, it also aims to vary the number of levels within a block as much as possible. If you really only want to block an orthogonal design, then you can perhaps use an efficient and orthogonal designs (so both the ;eff and ;orth command in your syntax) and block it. However, there is really nothing stopping you from blocking an efficient design, it will usually not lead to any problems. You can check the blocks manually if the levels within an attribute are varied enough within a block.

Mike

Hey Mike,

Many thanks for this info. Clear as crystal.
Hope you are well

Cheers

Tracey