[R-sig-DCM] Any package which can handle choice based sampling

Chris Chapman cnchapman at gmail.com
Tue May 2 16:12:42 CEST 2017


Not sure -- personally don't know of an existing function for such models,
but you could write your own likelihood function. That should work with
bayesm (see the reference book), and is documented for RSGHB:
https://www.rdocumentation.org/packages/RSGHB/versions/1.1.2/topics/doHB

The main question would be confidence about correctness of both the custom
likelihood function and in assumptions of the estimator itself (e.g.,
whether it's robust for an arbitrary likelihood landscape).

On Mon, May 1, 2017 at 8:34 PM, ravibabu manchala <
ravibabumanchala at gmail.com> wrote:

> Dear Cris
>
> Thanks for the response and thanks for the link. For a newcomer to R it
> has some very good material.
>
>
> I am aware of some of the models and that they can perform for Choice
> based conjoint (CBC) but I am not looking for CBC. I am looking at a case
> where sampling is not random but samples are decided as per the choices.
>
> I will clarify with an example. Suppose A, B and C are three choices in a
> universal set with shares of 75%,  24% and 1% respectively. As it is
> difficult to get the samples for C if you go for random sample you
> concentrate and get more samples for C. Say our sample proportions are 60%,
> 30% and 10% for A, B and C respectively. There are estimators designed for
> such samples (pl. see https://eml.berkeley.edu/~mcfadden/discrete/ch2.pdf
> for a review).
>
> To the extent I have reviewed none the packages are designed to estimate
> with choice based samples.
>
> May be I am missing something.
>
> with warm regards
> Ravibabu
>
>
> On 1 May 2017 at 22:32, Chris Chapman <cnchapman at gmail.com> wrote:
>
>> What exactly are you trying to do?
>>
>> Among the R packages that can handle estimation of choice models are
>> mlogit, bayesm, RSGHB, and ChoiceModelR. Choice-based conjoint models can
>> be fit by any of those. (For example, see Chapter 13 here:
>> http://r-marketing.r-forge.r-project.org/Instructor/slides-index.html)
>>
>>
>> On Mon, May 1, 2017 at 11:48 AM, ravibabu manchala <
>> ravibabumanchala at gmail.com> wrote:
>>
>>> Hello all,
>>>
>>> I am new to the group. I am looking for a choice based sampling estimator
>>> in R. Is there any package in R which handles this?
>>>
>>> Any leads would be most welcome. Thanks a lot in advance.
>>>
>>> regards
>>> Ravibabu
>>>
>>>         [[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> R-SIG-DCM mailing list
>>> R-SIG-DCM at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-dcm
>>>
>>
>>
>

	[[alternative HTML version deleted]]



More information about the R-SIG-DCM mailing list