|
|
| Line 1: |
Line 1: |
| {{regression bar}}
| | Oscar is what my wife loves to contact me and I completely dig that title. Her family members life in Minnesota. Bookkeeping is what I do. Doing ceramics is what love doing.<br><br>my blog; [http://dtekorea.com/xe/DS_030111/1073195 dtekorea.com] |
| In [[economics]], '''discrete choice''' models, or '''qualitative choice models''', describe, explain, and predict choices between two or more discrete alternatives, such as entering or not entering the [[labor market]], or choosing between modes of [[transport]]. Such choices contrast with standard consumption models in which the quantity of each good consumed is assumed to be a continuous variable. In the continuous case, calculus methods (e.g. first-order conditions) can be used to determine the optimum amount chosen, and demand can be modeled empirically using [[regression analysis]]. On the other hand, discrete choice analysis examines situations in which the potential outcomes are discrete, such that the optimum is not characterized by standard first-order conditions. Thus, instead of examining “how much” as in problems with continuous choice variables, discrete choice analysis examines “which one.” However, discrete choice analysis can also be used to examine the chosen quantity when only a few distinct quantities must be chosen from, such as the number of vehicles a household chooses to own <ref name="cars">[[Kenneth E. Train|Train, K.]] (1986). Qualitative Choice Analysis: Theory, Econometrics, and an Application to Automobile Demand, MIT Press, [http://emlab.berkeley.edu/books/choice.html Chapter 8].</ref>
| |
| and the number of minutes of telecommunications service a customer decides to purchase.<ref>[[Kenneth E. Train|Train, K.]] (1987). [[Daniel McFadden|McFadden, D.]], and Ben-Akiva, M., “[http://elsa.berkeley.edu/reprints/misc/demand.pdf The Demand for Local Telephone Service: A Fully Discrete Model of Residential Call Patterns and Service Choice], Rand Journal of Economics, Vol. 18, No. 1, pp109-123</ref> Techniques such as [[logistic regression]] and [[probit regression]] can be used for empirical analysis of discrete choice.
| |
| | |
| Discrete choice models theoretically or empirically model choices made by people among a finite set of alternatives. The models have been used to examine, e.g., the choice of which car to buy,<ref name="cars" /><ref>[[Kenneth E. Train|Train, K.]] and [http://www.brookings.edu/experts/w/winstonc.aspx/ Winston, C.] (2007). “[http://elsa.berkeley.edu/~train/tw104.pdf Vehicle Choice Behavior and the Declining Market Share of US Automakers],” International Economic Review, Vol. 48, No. 4, pp. 1469-1496.</ref>
| |
| where to go to college,<ref name="college">Fuller, WC, [[Charles F. Manski|Manski, C.]], and Wise, D. (1982). "[http://www.jstor.org/pss/145612 New Evidence on the Economic Determinants of Post-secondary Schooling Choices]." Journal of Human Resources 17(4): 477-498.</ref>
| |
| , which mode of [[transport]] (car, bus, rail) to take to work<ref name="bart">[[Kenneth E. Train|Train, K.]] (1978). “[http://elsa.berkeley.edu/~train/valtrb.pdf A Validation Test of a Disaggregate Mode Choice Model]”, Transportation Research, Vol. 12, pp. 167-174.</ref>
| |
| among numerous other applications. Discrete choice models are also used to examine choices by organizations, such as firms or government agencies. In the discussion below, the decision-making unit is assumed to be a person, though the concepts are applicable more generally. [[Daniel McFadden]] won the [[Nobel Memorial Prize in Economic Sciences|Nobel prize]] in 2000 for his pioneering work in developing the theoretical basis for discrete choice.
| |
| | |
| Discrete choice models statistically relate the choice made by each person to the attributes of the person and the attributes of the alternatives available to the person. For example, the choice of which car a person buys is statistically related to the person’s income and age as well as to price, fuel efficiency, size, and other attributes of each available car. The models estimate the probability that a person chooses a particular alternative. The models are often used to forecast how people’s choices will change under changes in demographics and/or attributes of the alternatives.
| |
| | |
| == Applications ==
| |
| | |
| * Marketing researchers use discrete choice models to study [[Consumer theory|consumer demand]] and to predict competitive business responses, enabling choice modelers to solve a range of business problems, such as [[pricing]], [[New product development|product development]], and [[Demand curve|demand estimation]] problems.<ref name="cars"/>
| |
| * Transportation planners use discrete choice models to predict demand for planned [[transport]]ation systems, such as which route a driver will take and whether someone will take [[rapid transit]] systems.<ref name="bart"/><ref>Ramming, M.S. (2001). “[http://library.mit.edu/item/001107149 Network Knowledge and Route Choice]”. Unpublished Ph.D. Thesis, Massachusetts Institute of Technology. MIT catalogue</ref> The first applications of discrete choice models were in transportation planning, and much of the most advanced research in discrete choice models is conducted by transportation researchers.
| |
| * Energy forecasters and policymakers use discrete choice models for households’ and firms’ choice of heating system, appliance efficiency levels, and fuel efficiency level of vehicles.<ref>Andrew Goett, Kathleen Hudson, and [[Kenneth E. Train|Train, K]] (2002). "Customer Choice Among Retail Energy Suppliers," Energy Journal, Vol. 21, No. 4, pp. 1-28.</ref><ref name="rt">David Revelt and [[Kenneth E. Train|Train, K]] (1998). "[http://www.jstor.org/stable/pdfplus/2646846.pdf Mixed Logit with Repeated Choices: Households' Choices of Appliance Efficiency Level]," Review of Economics and Statistics, Vol. 80, No. 4, pp. 647-657</ref>
| |
| * Environmental studies utilize discrete choice models to examine the recreators’ choice of, e.g., fishing or skiing site and to infer the value of amenities, such as campgrounds, fish stock, and warming huts, and to estimate the value of water quality improvements.<ref name="rec">[[Kenneth E. Train|Train, K]] (1998)."[http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.27.4879 Recreation Demand Models with Taste Variation]," Land Economics, Vol. 74, No. 2, pp. 230-239.</ref>
| |
| * Labor economists use discrete choice models to examine participation in the work force, occupation choice, and choice of college and training programs.<ref name="college"/>
| |
| * Evacuation modelling utilizes these models in order to simulate human behaviour during emergency situations.<ref name="rino">[[Lovreglio R., Borri D., dell'Olio L., Ibeas, A.]] (2013)."[http://dx.doi.org/10.1016/j.ssci.2013.10.004 A discrete choice model based on random utilities for exit choice in emergency evacuations]," Safety Science, Volume 62, February 2014, Pages 418–426</ref>
| |
| | |
| == Common Features of Discrete Choice Models ==
| |
| | |
| Discrete choice models take many forms, including: Binary Logit, Binary Probit, Multinomial Logit, Conditional Logit, Multinomial Probit, Nested Logit, Generalized Extreme Value Models, Mixed Logit, and Exploded Logit. All of these models have the features described below in common.
| |
| | |
| === Choice Set ===
| |
| | |
| The choice set is the set of alternatives that are available to the person. For a discrete choice model, the choice set must meet three requirements:
| |
| | |
| # The set of alternatives must be ''exhaustive'', meaning that the set includes all possible alternatives. This requirement implies that the person necessarily does choose an alternative from the set.
| |
| # The alternatives must be ''mutually exclusive'', meaning that choosing one alternative means not choosing any other alternatives. This requirement implies that the person chooses only one alternative from the set.
| |
| # The set must contain a ''finite'' number of alternatives. This third requirement distinguishes discrete choice analysis from forms of regression analysis in which the dependent variable can (theoretically) take an infinite number of values.
| |
| | |
| As an example, the choice set for a person deciding which mode of [[transport]] to take to work includes driving alone, carpooling, taking bus, etc. The choice set is complicated by the fact that a person can use multiple modes for a given trip, such as driving a car to a train station and then taking train to work. In this case, the choice set can include each possible combination of modes. Alternatively, the choice can be defined as the choice of “primary” mode, with the set consisting of car, bus, rail, and other (e.g. walking, bicycles, etc.). Note that the alternative “other” is included in order to make the choice set exhaustive.
| |
| | |
| Different people may have different choice sets, depending on their circumstances. For instance, the [[Scion (automobile)|Scion]] automobile was not sold in Canada as of 2009, so new car buyers in Canada faced different choice sets from those of American consumers. Such considerations are taken into account in the formulation of discrete choice models.
| |
| | |
| === Defining Choice Probabilities ===
| |
| | |
| A discrete choice model specifies the probability that a person chooses a particular alternative, with the probability expressed as a function of observed variables that relate to the alternatives and the person. In its general form, the probability that person ''n'' chooses alternative ''i'' is expressed as:
| |
| | |
| : <math> P_{ni} \equiv Prob( \text{Person } n \text{ chooses alternative } i) = G(x_{ni}, \;x_{nj}, \; j \neq i,\; s_n, \;\beta), </math>
| |
| where
| |
| | |
| : <math> \scriptstyle x_{ni} </math> is a vector of attributes of alternative ''i'' faced by person ''n'',
| |
| | |
| : <math> \scriptstyle x_{nj}, \; j \neq i </math> is a vector of attributes of the other alternatives (other than ''i'') faced by person ''n'',
| |
| | |
| : <math> s_n </math> is a vector of characteristics of person ''n'', and
| |
| | |
| : <math> \beta </math> is a set of parameters giving the effects of variables on probabilities, which are estimated statistically.
| |
| | |
| In the mode of [[transport]] example above, the attributes of modes (''x<sub>ni</sub>''), such as travel time and cost, and the characteristics of consumer (''s<sub>n</sub>''), such as annual income, age, and gender, can be used to calculate choice probabilities. The attributes of the alternatives can differ over people; e.g., cost and time for travel to work by car, bus, and rail are different for each person depending on the location of home and work of that person.
| |
| | |
| '''Properties:'''
| |
| * ''P<sub>ni</sub>'' is between 0 and 1
| |
| * <math> \scriptstyle \forall n:\; \sum_{j=1}^J P_{nj} = 1 , </math> where ''J'' is the total number of alternatives.
| |
| * (Expected fraction of people choosing ''i'' ) <math> = {1 \over N} {\sum_{n=1}^N P_{ni}}, </math> where N is the number of people making the choice.
| |
| | |
| Different models (i.e., models using a different function G) have different properties. Prominent models are introduced below.
| |
| | |
| === Consumer Utility ===
| |
| Discrete choice models can be derived from [[utility theory]]. This derivation is useful for three reasons:
| |
| | |
| # It gives a precise meaning to the probabilities ''P<sub>ni</sub>''
| |
| # It motivates and distinguishes alternative model specifications, e.g., the choice of a functional form for ''G''.
| |
| # It provides the theoretical basis for calculation of changes in consumer surplus (compensating variation) from changes in the attributes of the alternatives.
| |
| | |
| ''U<sub>ni</sub>'' is the utility (or net benefit or well-being) that person ''n'' obtains from choosing alternative ''i''. The behavior of the person is utility-maximizing: person n chooses the alternative that provides the highest utility. The choice of the person is designated by dummy variables, ''y<sub>ni</sub>'', for each alternative:
| |
| | |
| : <math> y_{ni} = \begin{cases}
| |
| 1, & if \, U_{ni} > U_{nj} , \quad j \not= i,\\
| |
| 0, & otherwise\end{cases}</math>
| |
| | |
| Consider now the researcher who is examining the choice. The person’s choice depends on many factors, some of which the researcher observes and some of which the researcher does not. The utility that the person obtains from choosing an alternative is decomposed into a part that depends on variables that the researcher observes and a part that depends on variables that the researcher does not observe. In a linear form, this decomposition is expressed as
| |
| : <math> U_{ni}= \beta z_{ni} + \varepsilon_{ni} </math>
| |
| | |
| where
| |
| | |
| : <math> \textstyle z_{ni} </math> is a vector of observed variables relating to alternative ''i'' for person ''n'' that depends on attributes of the alternative, ''x<sub>ni</sub>'', interacted perhaps with attributes of the person, ''s<sub>n</sub>'', such that it can be expressed as
| |
| ::: <math> \textstyle z_{ni}=z(x_{ni}, \, s_n) </math> for some numerical function ''z'',
| |
| : <math> \textstyle \beta </math> is a corresponding vector of coefficients of the observed variables, and
| |
| : <math> \varepsilon_{ni} </math> captures the impact of all unobserved factors that affect the person’s choice.
| |
| | |
| The choice probability is then
| |
| :<math>
| |
| \begin{align}
| |
| P_{ni}& = Prob(\, y_{ni} = 1 \,) = Prob(\, U_{ni} > U_{nj}, \quad j \not= i \,) \\
| |
| & = Prob(\, \beta z_{ni} + \varepsilon_{ni} > \beta z_{nj} + \varepsilon_{nj}, \; j \neq i \,) \\
| |
| & = Prob(\, \varepsilon_{nj}- \varepsilon_{ni} < \beta z_{ni}- \beta z_{nj}, \; j \neq i \,)
| |
| \end{align}
| |
| </math>
| |
| | |
| Given ''β'', the choice probability is the probability that the random terms, {{nowrap|''ε<sub>nj</sub>'' − ''ε<sub>ni</sub>''}} (which are random from the researcher’s perspective, since the researcher does not observe them) are below the respective quantities <math> \textstyle \forall j \neq i: \beta z_{ni} - \beta z_{nj}, \; </math>. Different choice models
| |
| (i.e. different specifications of G) arise from different distributions of ''ε<sub>ni</sub>'' for all ''i'' and different treatments of ''β''.
| |
| | |
| === Properties of Discrete Choice Models Implied by Utility Theory ===
| |
| ==== Only differences matter ====
| |
| | |
| The probability that a person chooses a particular alternative is determined by comparing the utility of choosing that alternative to the utility of choosing other alternatives:
| |
| | |
| :<math>
| |
| \begin{align}
| |
| P_{ni}& = Prob(\, y_{ni} = 1 \,) \\
| |
| & = Prob(\, U_{ni} > U_{nj}, \quad\forall j \not= i \,) \\
| |
| & = Prob(\, U_{ni} \, - \, U_{nj} > 0, \quad\forall j \not= i \,)
| |
| \end{align}
| |
| </math>
| |
| | |
| As the last term indicates, the choice probability depends only on the difference in utilities between alternatives, not on the absolute level of utilities. Equivalently, adding a constant to the utilities of all the alternatives does not change the choice probabilities.
| |
| | |
| ==== Scale must be normalized ====
| |
| | |
| Since utility has no units, it is necessary to normalize the scale of utilities. The scale of utility is often defined by the variance of the error term in discrete choice models. This variance may differ depending on the characteristics of the dataset, such as when or where the data are collected. Normalization of the variance therefore affects the interpretation of parameters estimated across diverse datasets.
| |
| | |
| == Prominent Types of Discrete Choice Models ==
| |
| Discrete choice models can first be classified according to the number of available alternatives.
| |
| : * Binomial choice models (dichotomous): 2 available alternatives
| |
| : * Multinomial choice models ([[polytomous choice|polytomous]]): 3 or more available alternatives
| |
| | |
| Multinomial choice models can further be classified according to the model specification:
| |
| : * Models, such as standard logit, that assume no correlation in unobserved factors over alternatives
| |
| : * Models that allow correlation in unobserved factors among alternatives
| |
| | |
| In addition, specific forms of the models are available for examining rankings of alternatives (i.e., first choice, second choice, third choice, etc.) and for ratings data.
| |
| | |
| Details for each model are provided in the following sections.
| |
| | |
| === Binary Choice ===
| |
| ===={{anchor|basic logit}} A. Logit with attributes of the person but no attributes of the alternatives ====
| |
| | |
| {{main|Logistic regression}}
| |
| | |
| ''U<sub>n</sub>'' is the utility (or net benefit) that person n obtains from taking an action (as opposed to not taking the action). The utility the person obtains from taking the action depends on the characteristics of the person, some of which are observed by the researcher and some are not:
| |
| : <math> U_n = \beta s_n + \varepsilon_n </math>
| |
| | |
| The person takes the action, {{nowrap|''y<sub>n</sub>'' {{=}} 1}}, if ''U<sub>n</sub>'' > 0. The unobserved term, ''ε<sub>n</sub>'', is assumed to have a [[logistic distribution]].
| |
| | |
| The specification is written succinctly as:
| |
| ** {{nowrap|''U<sub>n</sub>'' {{=}} ''βs<sub>n</sub>'' + ''ε<sub>n</sub>''}}
| |
| ** <math> y_n = \begin{cases}
| |
| 1, & if \, U_n > 0, \\
| |
| 0, & if \, U_n \le 0
| |
| \end{cases}</math>
| |
| ** {{nowrap|''ε'' ∼ }} [[Logistic distribution|Logistic]],
| |
| Then the probability of taking the action is
| |
| :: <math> Prob(y_n=1) = {1 \over 1+exp(-\beta s_n)} </math>
| |
| | |
| ===={{anchor|basic probit}} B. Probit with attributes of the person but no attributes of the alternatives ====
| |
| | |
| {{Main|Probit model}}
| |
| | |
| The description of the model is the same as [[#A. Logit with attributes of the person but no attributes of the alternatives|model '''A''']], except the unobserved terms are distributed [[Normal distribution|standard normal]] instead of [[Logistic function|logistic]].
| |
| | |
| ** {{nowrap|''U<sub>n</sub>'' {{=}} ''βs<sub>n</sub>'' + ''ε<sub>n</sub>''}}
| |
| ** <math> y_n = \begin{cases}
| |
| 1, & if \, U_n > 0, \\
| |
| 0, & if \, U_n \le 0
| |
| \end{cases}</math>
| |
| ** {{nowrap|''ε'' ∼ }} [[Normal distribution|Standard normal]],
| |
| Then the probability of taking the action is
| |
| :: <math> Prob(y_n=1) = \textstyle \Phi(\beta s_n) </math>,
| |
| :: where Φ() is [[Cumulative normal|cumulative distribution function of standard normal]].
| |
| | |
| ===={{anchor|logit varying over alternatives}} C. Logit with variables that vary over alternatives====
| |
| | |
| ''U<sub>ni</sub>'' is the utility person ''n'' obtains from choosing alternative ''i''. The utility of each alternative depends on the attributes of the alternatives interacted perhaps with the attributes of the person. The unobserved terms are assumed to have an [[Extreme value distribution|extreme value]] distribution.<ref group ="nb" name="ev" >
| |
| The density of the extreme value distribution is {{nowrap|''ƒ''(''ε<sub>nj</sub>'') {{=}} ''exp''( − ''ε<sub>nj</sub>'')''exp''( − ''exp''( − ''ε<sub>nj</sub>''))}}, and the cumulative distribution function is {{nowrap|''F''(''ε<sub>nj</sub>'') {{=}} ''exp''( − ''exp''( − ''ε<sub>nj</sub>'')).}}
| |
| | |
| This distribution is also called the [[Gumbel distribution|Gumbel]] or type I extreme value distribution, a special type of [[Extreme value distribution|generalized extreme value distribution]].
| |
| </ref>
| |
| | |
| ** {{nowrap|''U''<sub>''n''1</sub> {{=}} ''βz''<sub>''n''1</sub> + ''ε''<sub>''n''1</sub>, }}
| |
| ** {{nowrap|''U''<sub>''n''2</sub> {{=}} ''βz''<sub>''n''2</sub> + ''ε''<sub>''n''2</sub>, }}
| |
| ** <math> \varepsilon_{n1}, \; \varepsilon_{n2} \sim </math> [[iid]] [[Extreme value distribution|extreme value]],
| |
| which gives this expression for the probability
| |
| :: <math>
| |
| P_{n1}={exp(\beta z_{n1}) \over (exp(\beta z_{n1})+exp(\beta z_{n2}))}
| |
| </math>
| |
| | |
| We can relate this specification to [[#A. Logit with attributes of the person but no attributes of the alternatives|model ''' A ''']] above, which is also binary logit. In particular, ''P''<sub>''n''1</sub> can also be expressed as
| |
| :: <math>
| |
| P_{n1} = {1 \over (1+exp(-\beta (z_{n1}-z_{n2}))}
| |
| </math>
| |
| | |
| Note that if two error terms are [[iid]] [[Extreme value distribution|extreme value]],<ref group ="nb" name="ev" /> their difference is distributed [[Logistic function|logistic]], which is the basis for the equivalence of the two specifications.
| |
| | |
| ===={{anchor|probit varying over alternatives}} D. Probit with variables that vary over alternatives ====
| |
| The description of the model is the same as [[#C. Logit with variables that vary over alternatives|model '''C''']], except the difference of the two unobserved terms are distributed [[Normal distribution|standard normal]] instead of [[Logistic function|logistic]].
| |
| | |
| Then the probability of taking the action is
| |
| :: <math>
| |
| P_{n1} = \textstyle\Phi(\beta (z_{n1}-z_{n2})),
| |
| </math>
| |
| :: where Φ is the [[Cumulative normal|cumulative distribution function of standard normal]].
| |
| | |
| === Multinomial Choice without Correlation Among Alternatives ===
| |
| ===={{anchor|multinomial logit}} E. Logit with attributes of the person but no attributes of the alternatives ====
| |
| {{Main|Multinomial logit}}
| |
| | |
| The utility for all alternatives depends on the same variables, ''s<sub>n</sub>'', but the coefficients are different for different alternatives:
| |
| | |
| ** {{nowrap|''U<sub>ni</sub>'' {{=}} ''β<sub>i</sub>''''s<sub>n</sub>'' + ''ε<sub>ni</sub>'', }}
| |
| ** Since only differences in utility matter, it is necessary to normalize <math> \scriptstyle \beta_i =0 </math> for one alternative. Assuming <math> \scriptstyle \beta_1=0 </math>,
| |
| ** {{nowrap|''ε<sub>ni</sub>'' ∼ }} [[iid]] [[Extreme value distribution|extreme value]] <ref group ="nb" name="ev" />
| |
| The choice probability takes the form
| |
| :: <math>
| |
| P_{ni}= {exp(\beta_i s_n) \over \sum_{j=1}^J exp(\beta_j s_n)},
| |
| </math>
| |
| :: where J is the total number of alternatives.
| |
| | |
| ===={{anchor|multinomial logit varying over alternatives|conditional logit}} F. Logit with variables that vary over alternatives (also called conditional logit) ====
| |
| | |
| The utility for each alternative depends on attributes of that alternative, interacted perhaps with attributes of the person:
| |
| ** {{nowrap|''U<sub>ni</sub>'' {{=}} ''βz<sub>ni</sub>'' + ''ε<sub>ni</sub>'', }}
| |
| ** {{nowrap|''ε<sub>ni</sub>'' ∼ }} [[iid]] [[Extreme value distribution|extreme value]],<ref group ="nb" name="ev" />
| |
| The choice probability takes the form
| |
| :: <math>
| |
| P_{ni} = {exp(\beta z_{ni}) \over \sum_{j=1}^J exp(\beta z_{nj})},
| |
| </math>
| |
| :: where J is the total number of alternatives.
| |
| | |
| Note that [[#E. Logit with attributes of the person but no attributes of the alternatives|model '''E''']] can be expressed in the same form as model '''F''' by appropriate respecification of variables.
| |
| | |
| ** Let <math> \scriptstyle d_j^k </math> be a dummy variable that identifies alternative k:
| |
| ::<math> \scriptstyle d_j^k = \begin{cases}
| |
| \scriptstyle 1, & \scriptstyle if \, j=k, \\
| |
| \scriptstyle 0, & \scriptstyle otherwise
| |
| \end{cases}</math>
| |
| ** Multiply ''s<sub>n</sub>'' from [[#E. Logit with attributes of the person but no attributes of the alternatives|model '''E''']] with each of these dummies: <math> \scriptstyle w_{nj}^k=s_n \, d_j^k </math>.
| |
| ** Then, model '''F''' is obtained by using <math> \scriptstyle z_{nj} = \{ w^1_{nj}, w^2_{nj}, \ldots, w^J_{nj} \} </math> and <math> \scriptstyle \beta = \{ \beta_1, \beta_2,\ldots,\beta_J \} </math>, where J is the number of alternatives.
| |
| | |
| ==={{anchor|multinomial correlated alternatives}} Multinomial Choice with Correlation Among Alternatives ===
| |
| | |
| A standard logit model is not always suitable, since it assumes that there is no correlation in unobserved factors over alternatives. This lack of correlation translates into a particular pattern of substitution among alternatives that might not always be realistic in a given situation. This pattern of substitution is often called the [[Independence of irrelevant alternatives|Independence of Irrelevant Alternatives (IIA) property]] of standard logit models. See the [[Independence of irrelevant alternatives#IIA in econometrics|Red Bus/Blue Bus]] example <ref name=benakiva-lerman-1985>[http://cee.mit.edu/ben-akiva/ Ben-Akiva, M] and [http://cee.mit.edu/lerman/ Lerman, S] (1985). "[http://mitpress.mit.edu/catalog/item/default.asp?tid=8271&ttype=2 Discrete Choice Analysis: Theory and Application to Travel Demand (Transportation Studies)]", Massachusetts: MIT Press.</ref> or path choice example.<ref name=benakiva-bierlaire-1999>[http://cee.mit.edu/ben-akiva/ M. Ben-Akiva] and [http://roso.epfl.ch/mbi/ M. Bierlaire] (1999). “[http://roso.epfl.ch/mbi/handbook-final.pdf Discrete Choice Methods and Their Applications to Short Term Travel Decisions],” In R.W. Hall (ed.), Handbook of Transportation Science.</ref> A number of models have been proposed to allow correlation over alternatives and more general substitution patterns:
| |
| | |
| * Nested Logit Model - Captures correlations between alternatives by partitioning the choice set into 'nests'
| |
| ** Cross-nested Logit model<ref>Vovsha, P. (1997). "[http://trb.metapress.com/content/l341607q38j850j7/ Application of Cross-Nested Logit Model to Mode Choice in Tel Aviv, Israel, Metropolitan Area]," Transportation Research Record, 1607.
| |
| </ref> (CNL) - Alternatives may belong to more than one nest
| |
| ** C-logit Model<ref>Cascetta, E., A. Nuzzolo, F. Russo, and A.Vitetta (1996). “[http://www2.informatik.hu-berlin.de/alkox/lehre/lvws0809/verkehr/logit.pdf A Modified Logit Route Choice Model Overcoming Path Overlapping Problems: Specification and Some Calibration Results for Interurban Networks].” In J.B. Lesort (ed.), Transportation and Traffic Theory. Proceedings from the Thirteenth International Symposium on Transportation and Traffic Theory, Lyon, France, Pergamon pp. 697–711.
| |
| .</ref> - Captures correlations between alternatives using 'commonality factor'
| |
| ** Paired Combinatorial Logit Model<ref>Chu, C. (1989). “A Paired Combinatorial Logit Model for Travel Demand Analysis.” In Proceedings of the 5th World Conference on Transportation Research, 4, Ventura, CA, pp. 295–309.</ref> - Suitable for route choice problems.
| |
| | |
| * Generalized Extreme Value Model<ref>[[Daniel McFadden|McFadden, D.]] (1978). “[http://cowles.econ.yale.edu/P/cd/d04b/d0477.pdf Modeling the Choice of Residential Location].” In A. Karlqvist et al. (eds.), Spatial Interaction Theory and Residential Location, North Holland, Amsterdam pp. 75–96</ref> - General class of model, derived from the random utility model<ref name=benakiva-bierlaire-1999/> to which multinomial logit and nested logit belong
| |
| | |
| * Conditional probit <ref>J. Hausman and D. Wise (1978). "A Conditional Probit Model for Qualitative Choice: Discrete Decisions Recognizing Interdependence and Heterogenous Preferences," Econometrica, Vol. 48, No. 2, pp. 403-426</ref><ref name="dca">[[Kenneth E. Train|Train, K]](2003). "[http://elsa.berkeley.edu/books/choice2.html Discrete Choice Methods with Simulation]", Massachusetts: Cambridge University Press.</ref>- Allows full covariance among alternatives using a joint normal distribution.
| |
| | |
| * [[Mixed logit]] <ref name="rt" /><ref name="rec" /><ref name="dca" />- Allows any form of correlation and substitution patterns.<ref name=mt-mnl>[[Daniel McFadden|McFadden, D.]] and [[Kenneth E. Train|Train, K.]] (2000). “[http://elsa.berkeley.edu/wp/mcfadden1198/mcfadden1198.pdf Mixed MNL Models for Discrete Response],” Journal of Applied Econometrics, Vol. 15, No. 5, pp. 447-470,</ref> When a mixed logit is with jointly normal random terms, the models is sometimes called "multinomial probit model with logit kernel"<ref name=benakiva-bierlaire-1999/.<ref>[http://cee.mit.edu/ben-akiva/ M. Ben-Akiva] and [http://www.ecn.ulaval.ca/no_cache/professeurs/fiche_de_professeurs/?tx_fsgprofs_pi1%5Bprof%5D=7&tx_fsgprofs_pi1%5BbackPid%5D=60 D. Bolduc] (1996). “[http://elsa.berkeley.edu/reprints/misc/multinomial.pdf Multinomial Probit with a Logit Kernel and a General Parametric Specification of the Covariance Structure].” Working Paper.</ref> Can be applied to route choice <ref>[http://www.technion.ac.il/~civil/bekhor/ Bekhor, S.], [http://cee.mit.edu/ben-akiva/ Ben-Akiva, M.], and M.S. Ramming (2002). “[http://trb.metapress.com/content/126847136p81w0p3/ Adaptation of Logit Kernel to Route Choice Situation].” Transportation Research Record, 1805, 78–85.</ref>
| |
| | |
| The following sections describe Nested Logit, GEV, Probit, and Mixed Logit models in detail.
| |
| | |
| ==== {{anchor|nested logit}} G. Nested Logit and Generalized Extreme Value (GEV) models ====
| |
| | |
| The model is the same as [[#F. Logit with variables that vary over alternatives (also called conditional logit)|model '''F''']] except that the unobserved component of utility is correlated over alternatives rather than being independent over alternatives.
| |
| | |
| ** {{nowrap|''U<sub>ni</sub>'' {{=}} ''βz<sub>ni</sub>'' + ''ε<sub>ni</sub>'', }}
| |
| ** The marginal distribution of each ''ε<sub>ni</sub>'' is [[Extreme value distribution|extreme value]],<ref group ="nb" name="ev" /> but their joint distribution allows correlation among them.
| |
| ** The probability takes many forms depending on the pattern of correlation that is specified. See [[Generalized extreme value distribution|Generalized Extreme Value]].
| |
| | |
| ==== {{anchor|multinomial probit}} H. Multinomial Probit ====
| |
| {{Main|Multinomial probit}}
| |
| | |
| The model is the same as [[#G. Nested Logit and Generalized Extreme Value (GEV models)|model '''G''']] except that the unobserved terms are distributed jointly [[Normal distribution|normal]], which allows any pattern of correlation and [[heteroscedasticity]]:
| |
| ** {{nowrap|''U<sub>ni</sub>'' {{=}} ''βz<sub>ni</sub>'' + ''ε<sub>ni</sub>'', }}
| |
| ** <math> \scriptstyle \varepsilon_n \equiv (\varepsilon_{n1},\ldots,\varepsilon_{nJ}) \sim N(0,\Omega) , </math>
| |
| The choice probability is
| |
| :: <math>
| |
| \begin{align}
| |
| P_{ni} & = Prob(\beta z_{ni}+\varepsilon_{ni} > \beta z_{nj} + \varepsilon_{nj}, \; \forall j \; \ne \; i) \\
| |
| & = \int I(\beta z_{ni}+\varepsilon_{ni} > \beta z_{nj} + \varepsilon_{nj}, \; \forall j \; \ne \; i) \; \phi(\varepsilon_n | \Omega) \;d \varepsilon_n,
| |
| \end{align}
| |
| </math>
| |
| ::: where <math> \scriptstyle \phi(\varepsilon_n | \Omega) </math> is the joint normal density with mean zero and covariance <math> \scriptstyle \Omega </math>.
| |
| ** The integral for this choice probability does not have a closed form, and so the probability is approximated by quadrature or [http://elsa.berkeley.edu/choice2/ch5.pdf simulation].
| |
| ** When <math> \scriptstyle \Omega </math> is the identity matrix (such that there is no correlation or [[heteroscedasticity]]), the model is called independent probit.
| |
| | |
| ===={{anchor|mixed logit}} I. Mixed Logit ====
| |
| {{Main|Mixed logit}}
| |
| | |
| Mixed Logit models have become increasingly popular in recent years for several reasons. First, the model allows ''β'' to be random in addition to ''ε''. The randomness in ''β'' accommodates random taste variation over people and correlation across alternatives that generates flexible substitution patterns. Second, the advent in simulation has made approximation of the model fairly easy. In addition, [[Daniel McFadden|McFadden]] and [[Kenneth E. Train|Train]]<ref name="mt-mnl" /> have shown that any true choice model can be approximated, to any degree of accuracy by a mixed logit with appropriate specification of explanatory variables and distribution of coefficients.
| |
| | |
| ** {{nowrap|''U<sub>ni</sub>'' {{=}} ''βz<sub>ni</sub>'' + ''ε<sub>ni</sub>'', }}
| |
| ** <math> \scriptstyle \beta\; \sim f(\beta | \theta) </math> for any distribution <math> \it f </math>, where <math> \scriptstyle \theta </math> is the set of distribution parameters (e.g. mean and variance) to be estimated,
| |
| ** {{nowrap|''ε<sub>ni</sub>'' ∼ }} [[iid]] [[Extreme value distribution|extreme value]],<ref group ="nb" name="ev" />
| |
| The choice probability is
| |
| ::<math>
| |
| P_{ni}= \int_\beta L_{ni} (\beta) \, f(\beta | \theta) \, d\beta,
| |
| </math>
| |
| :: where
| |
| ::<math> L_{ni} (\beta) = {exp(\beta z_{ni}) \over {\sum_{j=1}^J exp(\beta z_{nj})}}</math> is logit probability evaluated at <math> \scriptstyle \beta, </math>
| |
| ::<math> J </math> is the total number of alternatives.
| |
| The integral for this choice probability does not have a closed form, so the probability is approximated by [http://elsa.berkeley.edu/choice2/ch6.pdf simulation]. Also see [[Mixed logit]] for further details.
| |
| | |
| ===Model Applications===
| |
| | |
| The models described above are adapted to accommodate rankings and ratings data.
| |
| | |
| ====Ranking of Alternatives====
| |
| In many situations, a person's ranking of alternatives is observed, rather than just their chosen alternative. For example, a person who has bought a new car might be asked what he/she would have bought if that car was not offered, which provides information on the person's second choice in addition to their first choice. Or, in a survey, a respondent might be asked:
| |
| | |
| ::<u>Example</u>: Rank the following cell phone calling plans from your most preferred to your least preferred.
| |
| :: * $60 per month for unlimited anytime minutes, two-year contract with $100 early termination fee
| |
| :: * $30 per month for 400 anytime minutes, 3 cents per minute after 400 minutes, one-year contract with $125 early termination fee
| |
| :: * $35 per month for 500 anytime minutes, 3 cents per minute after 500 minutes, no contract or early termination fee
| |
| :: * $50 per month for 1000 anytime minutes, 5 cents per minute after 1000 minutes, two-year contract with $75 early termination fee
| |
| | |
| The models described above can be adapted to account for rankings beyond the first choice. The most prominent model for rankings data is the exploded logit and its mixed version.
| |
| | |
| ====={{anchor|exploded logit}} J. Exploded Logit =====
| |
| | |
| Under the same assumptions as for a standard logit ([[#F. Logit with variables that vary over alternatives (also called conditional logit)|model '''F''']]), the probability for a ranking of the alternatives is a product of standard logits. The model is called "exploded logit" because the choice situation that is usually represented as one logit formula for the chosen alternative is expanded ("exploded") to have a separate logit formula for each ranked alternative. The exploded logit model is the product of standard logit models with the choice set decreasing as each alternative is ranked and leaves the set of available choices in the subsequent choice.
| |
| | |
| Without loss of generality, the alternatives can be relabeled to represent the person's ranking, such that alternative 1 is the first choice, 2 the second choice, etc. The choice probability of ranking J alternatives as 1, 2, …, J is then
| |
| :: <math>
| |
| Prob(ranking \; 1, 2, \ldots , J) = {exp(\beta z_1) \over \sum_{j=1}^J exp(\beta z_{nj})} {exp(\beta z_2) \over \sum_{j=2}^J exp(\beta z_{nj})} \ldots {exp(\beta z_{J-1}) \over \sum_{j=J-1}^J exp(\beta z_{nj})}
| |
| </math>
| |
| | |
| As with standard logit, the exploded logit model assumes no correlation in unobserved factors over alternatives. The exploded logit can be generalized, in the same way as the standard logit is generalized, to accommodate correlations among alternatives and random taste variation. The "mixed exploded logit" model is obtained by probability of the ranking, given above, for ''L<sub>ni</sub>'' in the mixed logit model ([[#I. Mixed Logit|model '''I''']]).
| |
| | |
| This model is also known in [[econometrics]] as the ''rank ordered logit model'' and it was introduced in that field by Beggs, Cardell and Hausman in 1981<ref name = "bch">Beggs, S., Cardell, S., Hausman, J., 1981. Assessing the potential demand for electric cars. Journal of Econometrics 17 (1),
| |
| 1–19 (September).</ref>{{,}}.<ref name = "combes" /> One application is the Combes et alii paper explaining the ranking of candidates to become professor.<ref name = "combes">Pierre-Philippe Combes, Laurent Linnemer, Michael Visser, Publish or peer-rich? The role of skills and networks in hiring economics professors, Labour Economics, Volume 15, Issue 3, June 2008, Pages 423-441, ISSN 0927-5371, 10.1016/j.labeco.2007.04.003. (http://www.sciencedirect.com/science/article/pii/S0927537107000413)</ref> It is also known as Plackett–Luce model in biomedical literature.<ref name = "combes" />
| |
| | |
| ==== Ratings Data ====
| |
| | |
| In survey, respondents are often asked to give ratings, such as:
| |
| | |
| ::<u>Example</u>: Please give your rating of how well the President is doing.
| |
| :: 1: Very badly
| |
| :: 2: Badly
| |
| :: 3: Okay
| |
| :: 4: Well
| |
| :: 5: Very well
| |
| | |
| Or,
| |
| ::<u>Example</u>: On a 1-5 scale where 1 means disagree completely and 5 means agree completely, how much do you agree with the following statement. "The Federal government should do more to help people facing foreclosure on their homes."
| |
| | |
| A multinomial discrete-choice model can examine the responses to these questions ([[#G. Nested Logit and Generalized Extreme Value (GEV) models|model '''G''']], [[#H. Multinomial Probit|model '''H''']], [[#I. Mixed Logit|model '''I''']]). However, these models are derived under the concept that the respondent obtains some utility for each possible answer and gives the answer that provides the greatest utility. It might be more natural to think that the respondent has some latent measure or index associated with the question and answers in response to how high this measure is. Ordered logit and ordered probit models are derived under this concept.
| |
| | |
| ====={{anchor|ordered logit}} K. Ordered Logit =====
| |
| {{Main|Ordered logit}}
| |
| | |
| Let ''U<sub>n</sub>'' represent the strength of survey respondent ''n''’s feelings or opinion on the survey subject. Assume that there are cutoffs of the level of the opinion in choosing particular response. For instance, in the example of the helping people facing foreclosure, the person chooses
| |
| # 1, if ''U<sub>n</sub>'' < a
| |
| # 2, if a < ''U<sub>n</sub>'' < b
| |
| # 3, if b < ''U<sub>n</sub>'' < c
| |
| # 4, if c < ''U<sub>n</sub>'' < d
| |
| # 5, if ''U<sub>n</sub>'' > d,
| |
| for some real numbers ''a'', ''b'', ''c'', ''d''.
| |
| | |
| Defining <math> U_n = \beta z_n + \varepsilon, \; \varepsilon \sim</math> [[Logistic function|Logistic]], then the probability of each possible response is:
| |
| : <math>
| |
| \begin{align}
| |
| Prob(choosing \, 1)
| |
| & = Prob(U_n <a) \\
| |
| &= Prob(\varepsilon < a - \beta z_n) \\
| |
| & = {1 \over 1+exp(-(a - \beta z_n))}
| |
| \end{align}
| |
| </math>
| |
| | |
| : <math>
| |
| \begin{align}
| |
| Prob(choosing \, 2)
| |
| & = Prob(a < U_n < b) \\
| |
| &= Prob(a- \beta z_n < \varepsilon < b - \beta z_n) \\
| |
| & = {1 \over 1+exp(-(b - \beta z_n))} - {1 \over 1+exp(-(a - \beta z_n))}
| |
| \end{align}
| |
| </math>
| |
| | |
| and so on up to
| |
| | |
| : <math>
| |
| \begin{align}
| |
| Prob(choosing \, 5)
| |
| & = Prob(U_n > d) \\
| |
| &= Prob(\varepsilon > d - \beta z_n) \\
| |
| & = 1 - {1 \over 1+exp(-(d - \beta z_n))}
| |
| \end{align}
| |
| </math>
| |
| | |
| The parameters of the model are the coefficients ''β'' and the cut-off points {{nowrap|''a − d''}}, one of which must be normalized for identification. When there are only two possible responses, the ordered logit is the same a binary logit ([[#A. Logit with attributes of the person but no attributes of the alternatives|model '''A''']]), with one cut-off point normalized to zero.
| |
| | |
| ====={{anchor|ordered probit}} L. Ordered Probit =====
| |
| {{Main|Ordered probit}}
| |
| | |
| The description of the model is the same as [[#K. Ordered Logit|model '''K''']], except the unobserved terms are distributed [[Normal distribution|standard normal]] instead of [[Logistic function|logistic]].
| |
| | |
| Then the choice probabilities are
| |
| | |
| :* {{nowrap|''Prob''(''choosing'' 1) {{=}} Φ(''a − βz<sub>n</sub>''), }}
| |
| :* {{nowrap|''Prob''(''choosing'' 2) {{=}} Φ(''b − βz<sub>n</sub>'') − Φ(''a − βz<sub>n</sub>''), }}
| |
| and so on.
| |
| where Φ(.) is the [[Cumulative normal|cumulative distribution function of standard normal]].
| |
| | |
| == Textbooks for further reading ==
| |
| | |
| * {{cite book| last = McFadden| first = Daniel L.| authorlink = Daniel McFadden| year = 1984| title = Econometric analysis of qualitative response models| series = Handbook of Econometrics, Volume II| volume = Chapter 24| publisher = Elsevier Science Publishers BV}}
| |
| *[http://cee.mit.edu/ben-akiva/ Ben-Akiva, M] and [http://cee.mit.edu/lerman/ S. Lerman] (1985). [http://mitpress.mit.edu/catalog/item/default.asp?tid=8271&ttype=2 ''Discrete Choice Analysis: Theory and Application to Travel Demand''], MIT Press.
| |
| *[http://www.econ.usyd.edu.au/staff/davidh/ Hensher, D.], [http://www.econ.usyd.edu.au/staff/johnr/ J. Rose], and [http://pages.stern.nyu.edu/~wgreene/ W. Greene] (2005). [http://books.google.com/books?hl=en&lr=&id=8yZrtCCABAgC&oi=fnd&pg=PR17&dq=Applied+Choice+Analysis:+A+Primer&ots=RCKM2_nbA4&sig=tKOOWUvIF3QcF-z8vN0wyxR7_4w ''Applied Choice Analysis: A Primer''], Cambridge University Press.
| |
| *[[G. S. Maddala|Maddala, G.]] (1983). [http://books.google.com/books?hl=en&lr=&id=-Ji1ZaUg7gcC&oi=fnd&pg=PR11&dq=G.S.+Maddala,+Limited-dependent+and+Qualitative+Variables+in+Econometrics,+New+York+:+Cambridge+University+Press,+1983.+&ots=7d1s4GmQHK&sig=knQEH5Ew6d_T-OQTzYYetvoIaJo ''Limited-dependent and Qualitative Variables in Econometrics''], Cambridge University Press.
| |
| *[[Kenneth E. Train|Train, K.]] (2003, 2009). [http://elsa.berkeley.edu/books/choice2.html ''Discrete Choice Methods with Simulation''], Cambridge University Press.
| |
| | |
| == Notes ==
| |
| | |
| <references group = "nb" />
| |
| | |
| == References ==
| |
| | |
| {{reflist|2}}
| |
| | |
| {{DEFAULTSORT:Discrete Choice}}
| |
| [[Category:Choice modelling]]
| |
| [[Category:Statistical models]]
| |
| [[Category:Single-equation methods (econometrics)]]
| |
| [[Category:Simultaneous equation methods (econometrics)]]
| |
| [[Category:Economics terminology]]
| |