Date:         Mon, 8 Nov 1999 19:04:19 +0100
Reply-To:     Thomas Otter <Thomas.Otter@WU-WIEN.AC.AT>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Thomas Otter <Thomas.Otter@WU-WIEN.AC.AT>
Subject:      Re: Conjoint Analysis
In-Reply-To:  <D704484B2D68D2118403002048403247E3736F@exchange30.usaa001. com>
Content-Type: text/plain; charset="us-ascii"

At 09:26 08.11.99 -0600, you wrote:
>Hello,
>
>Could anyone please explain how the three different probability-of-choice
>models are calculated:
>a) The maximum utility model

Calculate total utilities for all profiles in the choice set to
be evaluated and assume that every individual deterministically chooses the
profile with maximum utility. Allocate 100% choice probability to this profile
and 0% percent to all other profiles. Note that in applied settings the
procedure
is not that deterministic as it may appear becaues estimated partworths
contain
error.

>b) The Bradley Terry-Luce model

Divide the total utility of profile i by the sum of all total utilities in
the choice set to estimate an indivual's choice probability for profile i.


>c) The Logit model

Similar to the BTL model but take exp(total utility of profile i) and
divide by the
sum of all exp(utility) in the choice set to estimate an indivual's choice
probability for profile i.


Regards,
Thomas Otter

>
>Any help would be greatly appreciated.
>
>Mei-chih Huang
>
>