choice-metrics.com

Dear Andrew, Michiel and John

I am working on an SP off RP survey with a within and between modes experiment. As I understand from the manual, to use the .ref and .piv I need to have and alternative fixed with the reference value and the other alternative(s) to be pivoted around it. However, I am not interested in showing the reference one without any changes. I wanted to know if this approach will work (it technically runs, but I want to be sure that I can use the results directly):

U(alt1) = B_Cost[-0.01 ] * CostRef [10] * Cost[0.9,1,1.1]+
B_Time[-0.01] * TimeRef[10] * Time[0.9,1,1.1] /
U(alt2) = B_Cost * CostRef * Cost+
B_Time * TimeRef * Time
$

Alternatively, I could generate a reference alternative and never show it in the survey, but I'm afraid it will lose design properties.

Similarly between mode experiment I would like to use a multplier for the alternative mode to generate a faster/expensive alternative:


U(Current) = B_Cost[-0.01 ] * CostRef [10] * Cost[0.9,1,1.1]+
B_Time[-0.01] * TimeRef[10] * Time[0.9,1,1.1] /
U(other) = B_Cost * CostRef * Cost * CostMult[1.2]+
B_Time * TimeRef * Time * TimeMult [0.8]

Thanks in advance!

Hi,

Yes that approach should work for relative pivots (of course it wont work for absolute pivots). It is clever

My way around this is a bit easier. I first generate a design with absolute values using utility functions such as below.

U(alt1) = B_Cost[-0.01 ] * Cost[9,10,11]+ B_Time[-0.01] * Time[9,10,11] /
U(alt2) = B_Cost * Cost+B_Time * Time

Afterwards, I divide all the levels to the reference levels to obtain the pivots. This gives exactly the same result without the need to create interaction effects.

Michiel