1
$\begingroup$

I am interested in predicting shopping behaviour in a shopping center. I have a database with the chosen alternatives (shop) and variables describing that alternative (like type and size) and the individual which made the choice (sex, age). What I would like to get is the influcence of these variables on the behaviour.

My question: Would it be possible to append for each chosen shop the not chosen alternatives and calculate the binomial logistic model? As an example:

| person.id | shop.id | chosen | type    | size | sex | age |
|         1 | A       | yes    | fashion |    3 | m   |  21 |
|         1 | B       | no     | grocery |    5 | m   |  21 |
|         1 | C       | no     | fashion |    1 | m   |  21 |
|         2 | B       | yes    | grocery |    5 | f   |  45 |
|         2 | A       | no     | fashion |    3 | f   |  45 |
|         2 | C       | no     | fasion  |    1 | f   |  45 |

I was first thinking about the multinomial logistic regression but I am actually only interested in the variable chosen which is dichotomous (yes/no).

$\endgroup$
1
  • $\begingroup$ You might look into discrete choice modeling. $\endgroup$
    – RioRaider
    Oct 28, 2012 at 3:28

1 Answer 1

0
$\begingroup$

Multinomial logistic is the right method, I think. But it seems like you might need a multilevel model because you have multiple items for each person and those choices won't be independent.

This is equivalent to running many dichotomous logistic regressions, one on each type of item.

$\endgroup$
1
  • $\begingroup$ Thank you very much. I went with discrete models and multinomial logistic. The book "Discrete Choice Methods with Simulation" by Kenneth Train from here elsa.berkeley.edu/books/choice2.html helped me a lot. $\endgroup$
    – mrrrau
    Nov 1, 2012 at 20:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.