Demographic vs. lifestyle variables
- From: "Dino Hsu" <dino_hsu_1019@xxxxxxxxx>
- Date: 9 Jan 2006 10:11:09 -0800
Hi,
I want to do list augmentation which combines data from transaction
database and sampled data from survey with a common membership ID
(think of as in direct selling industry) and make estimation of missing
non-sampled survey variables. Some people call this process as "data
mapping" from database variables to survey variables with the model
trained from the sampled complete data.
The survey data belong to demographic or lifestyle variables, and
transaction database data belong to purchase behaviour variables. It's
said that demographic variables are not suitable as mapping target
because they are generally more "reliable" than lifestyle or even
behaviour variables, although this makes sense, but I need more
rationale to support this concept. According to this rule, higher
reliability variables can predict (map) lower reliability ones, but not
vice versa, and in terms of reliability we have: demographic >
behaviour > lifestyle in general.
Any comments are highly appreciated.
.
- Prev by Date: Balanced incomplete block designs (BIBD)
- Next by Date: Re: Correlations btw mixed Categorical & Interval Data
- Previous by thread: Balanced incomplete block designs (BIBD)
- Next by thread: Re: McNemar's hypothesis test
- Index(es):
Relevant Pages
|
|