f**n
发帖数: 401 | 1 Typical discrete choice model works on data with complete set of choice.
In some cases, information on the choice set from which choices are made may
be missing or incomplete, and the sample may lack data on some of the
alternatives. In an extreme case, data may be only available for one subset
of the alternatives, as when a firm keeps files on only its own customers.
For such data it would appear to be impossible to estimate a choice model.
Is this true? | w********g
发帖数: 294 | 2 see BLP first. nothing new... | l****o
发帖数: 2909 | 3 不是还有其他的generalized linear model可以适合lz说的这种情况吗？cencored /
tranucated regression很常见啊。 | h****i
发帖数: 1674 | 4 not really need go to BLP
you can use aggregation model to model this kind of data, but the
aggregation bias will not be easily handled.
aggregation means group those elementary alternatives, you need some
aggregate information.
I don't think BLP will get rid of the problem caused by missing
information.
choice.
made may
subset
customers.
model.
【在 f**n 的大作中提到】
: Typical discrete choice model works on data with complete set of choice.
: In some cases, information on the choice set from which choices are made may
: be missing or incomplete, and the sample may lack data on some of the
: alternatives. In an extreme case, data may be only available for one subset
: of the alternatives, as when a firm keeps files on only its own customers.
: For such data it would appear to be impossible to estimate a choice model.
: Is this true?
|
|
