choice-metrics.com

Best professor,

I have an efficient design with three alternatives. alt1, alt2 and PE-PE (alt3).So, there is a dummy variable combo that has 9 values 1 till 9. Each represent an combination of an access alternative (BTM, bike, walking) and an egress alternative (BTM, Bike, walking).

My question is. With eight zero's, in the dummy coefficient, that means that the 9th value of the combo attribute is the reference right?

Of all the three alternatives, the alternative 3 is the base. So how will I inpret this when estimating the coefficients? because I actually want the PE-PE alternative as base and this combination should not be shown in alternatve 1 and alternative 2.


Design
;alts = alt1, alt2, PE_alt
;rows = 27
;block = 3
;eff = (mnl,d)

;cond:

if(alt1.combo1 = [5,6,8,9], alt1.Cost1 = 0),
if(alt2.combo2 = [5,6,8,9], alt2.Cost2 = 0),
if(alt1.combo1 = [1,2,3,4,7,10], alt1.Cost1 > 0),
if(alt2.combo2 = [1,2,3,4,7,10], alt2.Cost2 > 0)

;model:
U(alt1) = b_att * AT1[5, 10, 15] access time
+ b_ett * ET1[5, 10, 15] egress time
+ b_cost * Cost1[0,0.5, 1.5, 3.0]
+ b_eff.dummy[0|0] * Effort1[1,2,0]
+ b_safe.dummy[0|0] * Safety1[1,2,0]
+ b_combo.dummy[0|0|0|0|0|0|0|0] * combo1[1,2,3,4,5,6,7,8,9] /

U(alt2) = b_att * AT2[5, 10, 15]
+ b_ett * ET2[5, 10, 15]
+ b_cost * Cost2[0,0.5, 1.5, 3.0]
+ b_eff[0|0] * Effort2[1,2,0]
+ b_safe[0|0] * Safety2[1,2,0]
+ b_combo.dummy[0|0|0|0|0|0|0|0] * combo2[1,2,3,4,5,6,7,8,9] /

U(PE_alt) = ASC_PE
+ b_att_PE * AT3[5, 10, 15]
+ b_ett_PE * ET3[5, 10, 15]
+ b_cost_PE * Cost3[0.5, 1.5, 3.0]
+ b_eff_PE.dummy[0|0] * Effort3[1,2,0]
+ b_safe_PE.dummy[0|0] * Safety3[1,2,0]

Yes, the last (9th) level of combo is the reference. If you want another level to be the reference, then you can moved it last (e.g. b_combo.dummy[0|0|0|0|0|0|0|0] * combo1[1,2,3,4,6,7,8,9,5] uses level 5 as reference level).

I am not sure what you mean with alternative 3 is "the base"? Are you linking the dummy coded attribute combo to alternative 3? An attribute is not the same as an alternative and a reference level in dummy coding is not related to an alternative. I probably misunderstand your question.

Michiel