Page 1 of 1

estimation issue regardining the 'opt-out'

PostPosted: Thu Mar 13, 2014 10:41 pm
by Anat Tchetchik
Dear all,
I ran a labelled design with 3 alts. and an opt-out that states "not interested" with no further attributes.
When I run the data in Stata's MNL what values do I need to put into the opt-out row under the columns of the the attributes?
Zeros? missing values? (in that case Stata ignores this obs. anyway) or should I insert only the three alts. row and omit the opt-out row at all?

Best regards,
Anat

Re: estimation issue regardining the 'opt-out'

PostPosted: Mon Mar 17, 2014 3:04 pm
by Michiel Bliemer
This is a Stata question, not an Ngene question. I am afraid I am not a State user, so I have no idea what input Stata requires. Perhaps anyone else reading this forum?

Re: estimation issue regardining the 'opt-out'

PostPosted: Wed Mar 19, 2014 7:34 am
by Anat Tchetchik
Yes, I guessed it's a Stata question... but Stata's list not very responsive.. so I just hope someone will have an idea...
Thanks anyway,
Anat

Re: estimation issue regardining the 'opt-out'

PostPosted: Wed Mar 19, 2014 7:14 pm
by dsr
I guess it depends on what the opt out is -
ie what are people choosing between and what does the opt out represent?

Is it products and the opt out is a "none of these"?
In that case I would have zeros for attribute levels and ASC dummy coded to 1 for that option.

If its management options or policies, then the opt out might be the status quo (SQ).
In that case I would use SQ attribute levels and an ASC dummy coded to 1 for that option.

Dan