1
$\begingroup$

At every trial a person chooses one of several options (the number of options varies from trial to trial). Every option has several numerical attributes. Attributes are the same for every option but their values can differ. We need to determine how every attribute affects the person's choice.

For example, a customer chooses a supplier depending on the price and warranty period. One can expect that the price has negative and the warranty period has positive effect on the customer's choice.

Which statistical method fits best here?

$\endgroup$
1
  • $\begingroup$ The price could be a positive attribute if it is a bargain and the customer could be uncomfortable with warranty if it seems to be too short. $\endgroup$ Commented Apr 16, 2017 at 19:10

1 Answer 1

0
$\begingroup$

You are trying to model individuals' discrete choices. A typical approach would be to estimate a multinomial logit (MNL) model with the observed choices as binary (0/1) dependent variable and characteristics of the choice options (e.g., price, quality, etc.) as independent variables. With R, there is many good packages to model discrete choices: survival (clogit), mlogit, RSGHB, gmnl, etc.

$\endgroup$
1
  • $\begingroup$ Thanks @Umka. The problem is that in my data every observation has unique options. Do you know which package can deal with it? It seems that mlogit can't. $\endgroup$
    – 8k14
    Commented Jul 5, 2017 at 7:02

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.