We analyze continuous approximations of discrete choice models with a large number of options. We start with a discrete choice model where agents choose between different options, and where each option is defined by a characteristic vector and a utility level. For each option, the characteristic vector and the utility level are random and jointly dependent. We analyze the optimal choice, which we define as the characteristic vector of the option with the highest utility level. This optimal choice is a random variable. The continuous approximation of the discrete choice model is the distributional limit of this random variable as the number of offers tends to infinity. We use point process theory and extreme value theory to derive an analytic expression for the continuous approximation, and show that this can be done for a range of distributional assumptions. We illustrate the theory by applying it to commuting data. We also extend the initial results by showing how the theory works when characteristics belong to an infinite-dimensional space, and by proposing a setup which allows us to further relax our distributional assumptions.
|Original language||English (US)|
|Title of host publication||Stochastic Processes and Applications - SPAS2017|
|Editors||Sergei Silvestrov, Anatoliy Malyarenko, Milica Rančić|
|Publisher||Springer New York LLC|
|Number of pages||23|
|State||Published - 2018|
|Event||International Conference on “Stochastic Processes and Algebraic Structures – From Theory Towards Applications”, SPAS 2017 - Västerås and Stockholm, Sweden|
Duration: Oct 4 2017 → Oct 6 2017
|Name||Springer Proceedings in Mathematics and Statistics|
|Conference||International Conference on “Stochastic Processes and Algebraic Structures – From Theory Towards Applications”, SPAS 2017|
|City||Västerås and Stockholm|
|Period||10/4/17 → 10/6/17|
Bibliographical noteFunding Information:
The authors wish to thank Dmitrii Silvestrov for valuable comments on the manuscript. Ola Hössjer’s work was supported by the Swedish Research Council, contract nr. 621-2008-4946, and the Gustafsson Foundation for Research in Natural Sciences and Medicine.
Acknowledgements The authors wish to thank Dmitrii Silvestrov for valuable comments on the manuscript. Ola Hössjer’s work was supported by the Swedish Research Council, contract nr. 621-2008-4946, and the Gustafsson Foundation for Research in Natural Sciences and Medicine.
© 2018, Springer Nature Switzerland AG.
- Concomitant of order statistics
- Discrete choice
- Extreme value theory
- Point processes
- Random fields
- Random utility