[Limdep Nlogit List] Help with clogit model
Vassilopoulos Achilleas
avassilopoulos.aua at gmail.com
Thu Apr 29 05:36:49 EST 2010
Manin,
I'm sorry but I misunderstood your first email. Now I see that what you have
created through SPSS is a design for conjoint analysis and not for a choice
experiment. So this is completely different from the design you would use
for a CE. Those 16 'profiles' you have now may only be used for ranking or
rating, and even in this case you should use all 16 together.
So my answer to your question is no; given the random partition of the cards
you have done, I am not sure ANY model can give you the desirable parameter
estimates; I can assure you though, that the conditional logit will not.
Good luck,
_____________ - _______________
Achilleas Vassilopoulos
Agricultural University of Athens,
Dept. of Agricultural Economics and Rural Development,
Lab. of Political Economy and European Integration.
Iera Odos 75, 11855, Athens, Greece
Tel: (+30) 210-5294726
Fax: (+30) 2105294786
-----Original Message-----
From: limdep-bounces at limdep.itls.usyd.edu.au
[mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of Manit
Satitsamitpong
Sent: Wednesday, April 28, 2010 6:05 PM
To: Limdep and Nlogit Mailing List
Subject: Re: [Limdep Nlogit List] Help with clogit model
Dear Professor Vassilopoulos,
Thank you very much for your detailed answer. The way I came up with the 16
alternatives is via SPSS orthogonal design command. Then I splitted into 4
choice sets to reduce the size of cards as you might have thought by
arbitrary process without using blocking technique. In this case, is there
any other ways I could measure the parameter estimates for each attribute?
Thank you very much for your help.
Best regards,
Manit Satitsamitpong
On Wed, Apr 28, 2010 at 11:03 PM, Achilleas Vassilopoulos <
avassilopoulos.aua at gmail.com> wrote:
> Dear Martin,
>
> When respondents choose between air, bus, train etc., we are talking about
> a
> labeled experiment whereas yours is an unlabeled one. You can easily
> estimate unlabeled choice experiments using Nlogit by just assigning
random
> names in each of your alternatives (e.g A, B, C, D). Note that although A,
> B, C and D will be different for every respondent, since you do not
include
> a constant term in your model (i.e "ONE" in your Rhs variables) the
results
> you get are the generic parameter estimates of the utility associated with
> each attribute or attribute level (i.e price, video, setup_cost,
> enhance_service).
>
> What I'm not sure about is how have you ended up with these 16
alternatives
> (combinations as you refer to)? Where they a result of a full or
> fractional
> factorial design or you have assigned the attributes and the attributes'
> level of each alternative arbitrarily or by experience? In the last
case,
> there is no possible way to estimate such a model in an efficient and
> unbiased manner since your Rhs variables are probably highly correlated
and
> the parameters you'll get (if any) will be completely uninformative. At
> last, instead of just splitting all 16 possible alternatives in 4 choice
> sets to reduce the size of cards, you need to employ blocking techniques
in
> your experimental design.
>
> I hope that helps,
> _____________ - _______________
>
> Achilleas Vassilopoulos
>
> Agricultural University of Athens,
> Dept. of Agricultural Economics and Rural Development,
> Lab. of Political Economy and European Integration.
> Iera Odos 75, 11855, Athens, Greece
>
> Tel: (+30) 210-5294726
> Fax: (+30) 2105294786
>
>
>
> -----Original Message-----
> From: limdep-bounces at limdep.itls.usyd.edu.au
> [mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of Manit
> Satitsamitpong
> Sent: Wednesday, April 28, 2010 3:21 PM
> To: limdep at limdep.itls.usyd.edu.au
> Subject: [Limdep Nlogit List] Help with clogit model
>
> Dear All,
>
> I am very new in this area. I really hope someone could help me with the
> command or concept about the choice model that I am facing right now.
>
>
> Suppose I would like to estimate customer preferences and the relative
> importance of service factors of IPTV such as price per month (range of
> price), video on demand feature (high or low), setup cost (high or low),
> enhanced service (yes or no),and other 2 factors. Then I created say 16
> profiles of combinations of those factors (conjoint questions) for
> respondents to answer. I understood that if I asked respondents to rank
> all
> those 16 profiles (combinations) then I can use SPSS to run the conjoint
> analysis and then I can calculate the importance of each factor.
>
> My question here is what if I break down those 16 profiles into 4 choices
> sets with each set contains 4 profiles (combinations) and then ask
> respondents to choose the best choice (alternative) among each choice set.
> In this case, I don't think I can use clogit command to do so. Because
> clogit command asks me to provide the choice list (specific names of each
> choice) but in this case I don't have a specific name for each combination
> of choice set.
>
> clogit ;lhs = choice
> ;Rhs = price, video, setup_cost, enhance_service
> ;Choices = ????
> $
>
> Here I don't know what to use as the parameters for Choices since all
those
> 16 combinations are different, unlike the transportation mode (air, bus,
> car, train) given in Professor Greene's book. Is there any way for me to
> use LIMDEP or other software to run this data given the data that i have
> (choice data instead of ranking data). Thank you very much. I am really
> appreciated your time for reading this mail.
>
>
> Best regards,
> Manit Satitsamitpong
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
>
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au
More information about the Limdep
mailing list