Q: Earlier when Rob asked why you included price in
the model you said there were 2 types of people--those that respond
to price more and the other, more independent kind, who don't care
and will pay whatever price. But your model is linear. Wouldn't a
mixture model be more appropriate and capture more of what you
care about.
A: Right, this is at the aggregate store level, so the difference
here is that your thinking it at this individual level and once
- what would happen is that - once you started aggregating these
people you don't have this well-defined extremes in the
population. What happens is, if you did that you would have this
bimodal distribution, but now you've got all these extremes some
people it's important some of the time and when you start
looking at it you still get some normal effects and when you
look at it at the aggregate level a lot of these individual
behaviors that explain how prices behave aren't applied at the
aggregate level. So there's this difference of aggregate and
individual level.
Q: If you look at customers in the store, whether
the distribution was bimodal or unimodal to reflect that individual
customers have different cutoffs: it does not seem to be quite
captured in your model.
A: I think it goes back to the other point, maybe my model is
inadequate, it's different to go back and say yes, there are a
lot of xxx possibilities for the model, but the point is, is
that the own price is the most important variable model and that
drives most of the action that you see in movement. There are a lot
of these other effects that are also important and I think it's
just more thinking about them as extensions, but the point is, I
think I've captured the essence to the problem, if you want to
start going beyond that, I think you're right. There are a lot
of reasons you might interested in some of these other effects,
but to get at those, you really would need a different type of
data set. The whole point of the way I phrased the problem -
let's think about the information set that's available to the
retailer. So all this stuff is internal data, something that
they've got easy access to and I think that's really what's
important, they're making the decisions.
Back