Wsrtjtyk

Part of a series on Statistics
Regression analysis
Models
Linear regression Simple regression Polynomial regression General linear model
Generalized linear model Discrete choice Logistic regression Multinomial logit Mixed logit Probit Multinomial probit Ordered logit Ordered probit Poisson
Multilevel model Fixed effects Random effects Mixed model
Nonlinear regression Nonparametric Semiparametric Robust Quantile Isotonic Principal components Least angle Local Segmented
Errors-in-variables
Estimation
Least squares Linear Non-linear
Ordinary Weighted Generalized
Partial Total Non-negative Ridge regression Regularized
Least absolute deviations Iteratively reweighted Bayesian Bayesian multivariate
Background
Regression model validation Mean and predicted response Errors and residuals Goodness of fit Studentized residual Gauss–Markov theorem
Statistics portal
v t e

In statistics and econometrics, the multivariate probit model is a generalization of the probit model used to estimate several correlated binary outcomes jointly. For example, if it is believed that the decisions of sending at least one child to public school and that of voting in favor of a school budget are correlated (both decisions are binary), then the multivariate probit model would be appropriate for jointly predicting these two choices on an individual-specific basis. This approach was initially developed by Siddhartha Chib and Edward Greenberg.^[1]

Example: bivariate probit

In the ordinary probit model, there is only one binary dependent variable Y{displaystyle Y} and so only one latent variable Y∗{displaystyle Y^{*}} is used. In contrast, in the bivariate probit model there are two binary dependent variables Y1{displaystyle Y_{1}} and Y2{displaystyle Y_{2}}, so there are two latent variables: Y1∗{displaystyle Y_{1}^{*}} and Y2∗{displaystyle Y_{2}^{*}}.
It is assumed that each observed variable takes on the value 1 if and only if its underlying continuous latent variable takes on a positive value:

Y1={1if Y1∗>0,0otherwise,{displaystyle Y_{1}={begin{cases}1&{text{if }}Y_{1}^{*}>0,\0&{text{otherwise}},end{cases}}}

Y2={1if Y2∗>0,0otherwise,{displaystyle Y_{2}={begin{cases}1&{text{if }}Y_{2}^{*}>0,\0&{text{otherwise}},end{cases}}}

with

{Y1∗=X1β1+ε1Y2∗=X2β2+ε2{displaystyle {begin{cases}Y_{1}^{*}=X_{1}beta _{1}+varepsilon _{1}\Y_{2}^{*}=X_{2}beta _{2}+varepsilon _{2}end{cases}}}

and

[ε1ε2]∣X∼N([00],[1ρρ1]){displaystyle {begin{bmatrix}varepsilon _{1}\varepsilon _{2}end{bmatrix}}mid Xsim {mathcal {N}}left({begin{bmatrix}0\0end{bmatrix}},{begin{bmatrix}1&rho \rho &1end{bmatrix}}right)}

Fitting the bivariate probit model involves estimating the values of β1, β2,{displaystyle beta _{1}, beta _{2},} and ρ{displaystyle rho }. To do so, the likelihood of the model has to be maximized. This likelihood is

L(β1,β2)=(∏P(Y1=1,Y2=1∣β1,β2)Y1Y2P(Y1=0,Y2=1∣β1,β2)(1−Y1)Y2P(Y1=1,Y2=0∣β1,β2)Y1(1−Y2)P(Y1=0,Y2=0∣β1,β2)(1−Y1)(1−Y2)){displaystyle {begin{aligned}L(beta _{1},beta _{2})={Big (}prod &P(Y_{1}=1,Y_{2}=1mid beta _{1},beta _{2})^{Y_{1}Y_{2}}P(Y_{1}=0,Y_{2}=1mid beta _{1},beta _{2})^{(1-Y_{1})Y_{2}}\[8pt]&{}qquad P(Y_{1}=1,Y_{2}=0mid beta _{1},beta _{2})^{Y_{1}(1-Y_{2})}P(Y_{1}=0,Y_{2}=0mid beta _{1},beta _{2})^{(1-Y_{1})(1-Y_{2})}{Big )}end{aligned}}}

Substituting the latent variables Y1∗{displaystyle Y_{1}^{*}} and Y2∗{displaystyle Y_{2}^{*}} in the probability functions and taking logs gives

∑(Y1Y2ln⁡P(ε1>−X1β1,ε2>−X2β2)+(1−Y1)Y2ln⁡P(ε1<−X1β1,ε2>−X2β2)+Y1(1−Y2)ln⁡P(ε1>−X1β1,ε2<−X2β2)+(1−Y1)(1−Y2)ln⁡P(ε1<−X1β1,ε2<−X2β2)).{displaystyle {begin{aligned}sum &{Big (}Y_{1}Y_{2}ln P(varepsilon _{1}>-X_{1}beta _{1},varepsilon _{2}>-X_{2}beta _{2})\[4pt]&{}quad {}+(1-Y_{1})Y_{2}ln P(varepsilon _{1}<-X_{1}beta _{1},varepsilon _{2}>-X_{2}beta _{2})\[4pt]&{}quad {}+Y_{1}(1-Y_{2})ln P(varepsilon _{1}>-X_{1}beta _{1},varepsilon _{2}<-X_{2}beta _{2})\[4pt]&{}quad {}+(1-Y_{1})(1-Y_{2})ln P(varepsilon _{1}<-X_{1}beta _{1},varepsilon _{2}<-X_{2}beta _{2}){Big )}.end{aligned}}}

After some rewriting, the log-likelihood function becomes:

∑(Y1Y2ln⁡Φ(X1β1,X2β2,ρ)+(1−Y1)Y2ln⁡Φ(−X1β1,X2β2,−ρ)+Y1(1−Y2)ln⁡Φ(X1β1,−X2β2,−ρ)+(1−Y1)(1−Y2)ln⁡Φ(−X1β1,−X2β2,ρ)).{displaystyle {begin{aligned}sum &{Big (}Y_{1}Y_{2}ln Phi (X_{1}beta _{1},X_{2}beta _{2},rho )\[4pt]&{}quad {}+(1-Y_{1})Y_{2}ln Phi (-X_{1}beta _{1},X_{2}beta _{2},-rho )\[4pt]&{}quad {}+Y_{1}(1-Y_{2})ln Phi (X_{1}beta _{1},-X_{2}beta _{2},-rho )\[4pt]&{}quad {}+(1-Y_{1})(1-Y_{2})ln Phi (-X_{1}beta _{1},-X_{2}beta _{2},rho ){Big )}.end{aligned}}}

Note that Φ{displaystyle Phi } is the cumulative distribution function of the bivariate normal distribution. Y1{displaystyle Y_{1}} and Y2{displaystyle Y_{2}} in the log-likelihood function are observed variables being equal to one or zero.

Multivariate Probit

For the general case, yi=(y1,...,yj), (i=1,...,N){displaystyle mathbf {y_{i}} =(y_{1},...,y_{j}), (i=1,...,N)} where we can take j{displaystyle j} as choices and i{displaystyle i} as individuals or observations, the probability of observing choice yi{displaystyle mathbf {y_{i}} } is

Pr(yi|Xiβ,Σ)=∫AJ⋯∫A1fN(yi∗|Xiβ,Σ)dy1∗…dyJ∗Pr(yi|Xiβ,Σ)=∫1y∗∈AfN(yi∗|Xiβ,Σ)dyi∗{displaystyle {begin{aligned}Pr(mathbf {y_{i}} |mathbf {X_{i}beta } ,Sigma )=&int _{A_{J}}cdots int _{A_{1}}f_{N}(mathbf {y} _{i}^{*}|mathbf {X_{i}beta } ,Sigma )dy_{1}^{*}dots dy_{J}^{*}\Pr(mathbf {y_{i}} |mathbf {X_{i}beta } ,Sigma )=&int mathbb {1} _{y^{*}in A}f_{N}(mathbf {y} _{i}^{*}|mathbf {X_{i}beta } ,Sigma )dmathbf {y} _{i}^{*}end{aligned}}}

Where A=A1×⋯×AJ{displaystyle A=A_{1}times cdots times A_{J}} and,

Aj={(−∞,0]yj∗=0(0,∞)yj∗=1{displaystyle A_{j}={begin{cases}(-infty ,0]&y_{j}^{*}=0\(0,infty )&y_{j}^{*}=1end{cases}}}

The log-likelihood function in this case would be
∑i=1Nlog⁡Pr(yi|Xiβ,Σ){displaystyle sum _{i=1}^{N}log Pr(mathbf {y_{i}} |mathbf {X_{i}beta } ,Sigma )}

Except for J≤2{displaystyle Jleq 2} typically there is no closed form solution to the integrals in the log-likelihood equation. Instead simulation methods can be used to simulated the choice probabilities. Methods using importance sampling include the GHK algorithm (Geweke, Hajivassilou, McFadden and Keane),^[2] AR (accept-reject), Stern's method. There are also MCMC approaches to this problem including CRB (Chib's method with Rao-Blackwellization), CRT (Chib, Ritter, Tanner), ARK (accept-reject kernel), and ASK (adaptive sampling kernel).^[3]. A variational approach scaling to large datasets is proposed in Probit-LMM (Mandt, Wenzel, Nakajima et al.).^[4]

References

Further reading

• Greene, William H., Econometric Analysis, seventh edition, Prentice-Hall, 2012.