Wyznaczanie funkcji wiarygodności dla IV-probit

Mam więc model binarny, w którym $y_1^*$ jest ukrytą nieobserwowaną zmienną, a $y_1 \in \{0,1\}$ jest obserwowany. $y_2$ określa $y_1$ a $z_2$ jest zatem moim instrumentem. Krótko mówiąc, model jest.

\begin{array}{rcl} y_{1}^{*} & = & δ_{1} z_{1} + α_{1} y_{2} + u_{1} \\ y_{2} & = & δ_{21} z_{1} + δ_{22} z_{2} + v_{2} = z δ + v_{2} \\ y_{1} & = & 1 [y^{*} > 0] \end{array}

$\begin{eqnarray} y_1^*&=& \delta_1 z_1 + \alpha_1 y_2 + u_1 \\ y_2 &=& \delta_{21} z_1 + \delta_{22}z_2 + v_2 = \textbf{z}\delta + v_2 \\ y_1 &=& \text{1}[y^*>0] \end{eqnarray}$ Ponieważ terminy błędów nie są niezależne, ale

\begin{array}{rcl} (\begin{matrix} u_{1} \\ v_{2} \end{matrix}) \sim N (0, [\begin{matrix} 1 & η \\ η & τ^{2} \end{matrix}]) . \end{array}

$\begin{eqnarray} \begin{pmatrix} u_1 \\ v_2 \end{pmatrix} \sim \mathcal{N} \left(\textbf{0} \; , \begin{bmatrix} 1 &\eta \\ \eta &\tau^2 \end{bmatrix} \right). \nonumber \end{eqnarray}$ Korzystam z modelu IV-probit.

Mam problem z uzyskaniem funkcji wiarygodności. Rozumiem, że mogę zapisać jeden z terminów błędów jako funkcję liniową drugiego, więc

\begin{array}{rcl} u_{1} = \frac{η}{τ^{2}} v_{2} + ξ, where ξ \sim N (0, 1 - η^{2}) . \end{array}

$\begin{eqnarray} u_{1} = \frac{\eta}{\tau^2}v_{2} + \xi, \qquad \text{where} \quad \xi \sim \mathcal{N}(0, 1-\eta^2). \end{eqnarray}$
i że

ξ

$\xi$ należy zastosować w celu nałożenia normalnego CDF.

Zajrzałem do instrukcji Stata ( http://www.stata.com/manuals13/rivprobit.pdf ) pod kątem IV-probit i sugerują użycie definicji gęstości warunkowej

\begin{array}{rcl} f (y_{1}, y_{2} ∣ z) = f (y_{1} ∣ y_{2}, z) f (y_{2} ∣ z) \end{array}

$\begin{eqnarray} f(y_1, y_2 \mid \textbf{z}) = f(y_1 \mid y_2, \textbf{z}) f(y_2 \mid \textbf{z}) \end{eqnarray}$
w celu uzyskania funkcji wiarygodności, ale tak naprawdę jej nie używam (i tak, mam w końcu zły wynik ...). Jak dotąd próbuję

Jak powiedziałem, nie użyłem definicji funkcji gęstości złącza, jak podano powyżej. Co więcej, kończę również na tym, żepodnosi się doco wydaje się błędne. Czy ktoś może mi podpowiedzieć, jak uzyskać prawidłową funkcję (log-) prawdopodobieństwa lub gdzie popełniłem błąd?

\begin{array}{rcl} L (y_{1}) & = & \prod_{i = 1}^{n} Pr (y_{1} = 0 ∣ y_{2}, z)^{1 - y_{1}} Pr (y_{1} = 1 ∣ y_{2}, z)^{y_{1}} \\ = & \prod_{i = 1}^{n} Pr (y_{1}^{*} \leq 0)^{1 - y_{1}} (Pr (y_{1}^{*} > 0) f (y_{2} ∣ z))^{y_{1}} \\ [standardizing] & = & \prod_{i = 1}^{n} Pr (\frac{ξ}{\sqrt{1 - η^{2}}} \leq - \frac{δ_{1} z_{1} + α_{1} y_{2} + \frac{η}{τ^{2}} (y_{2} - z)}{\sqrt{1 - η^{2}}})^{1 - y_{1}} \\ \cdot & (Pr (\frac{ξ}{\sqrt{1 - η^{2}}} < \frac{δ_{1} z_{1} + α_{1} y_{2} + \frac{η}{τ^{2}} (y_{2} - z)}{\sqrt{1 - η^{2}}}) f (y_{2} ∣ z))^{y_{1}} \\ = & [1 - Φ (w)]^{1 - y_{i}} {[Φ (w) f (y_{2} ∣ x)]}^{y_{1}} \end{array}

$\begin{eqnarray} \mathcal{L}(y_1) &=& \prod_{i=1}^n \Pr(y_1=0 \mid y_2, \textbf{z} )^{1-y_1} \Pr(y_1=1 \mid y_2, \textbf{z} )^{y_1} \nonumber \\ &=& \prod_{i=1}^n \Pr(y_1^* \leq 0)^{1-y_1} \Big(\Pr(y_1^* > 0) f(y_2 \mid \textbf{z}) \Big)^{y_1} \nonumber \\ \text{[standardizing]} &=& \prod_{i=1}^n \Pr \Big( \frac{\xi}{\sqrt{1-\eta^2}} \leq - \frac{\delta_1 z_1 + \alpha_1 y_2 + \frac{\eta}{\tau^2}(y_2 - \textbf{z})}{\sqrt{1-\eta^2}}\Big)^{1-y_1} \\ &\cdot& \Big(\Pr \Big( \frac{\xi}{\sqrt{1-\eta^2}} < \frac{\delta_1 z_1 + \alpha_1 y_2 + \frac{\eta}{\tau^2}(y_2 - \textbf{z})}{\sqrt{1-\eta^2}}\Big) f(y_2 \mid \textbf{z}) \Big)^{y_1} \nonumber \\ &=& [1-\Phi(w)]^{1-y_i} \left[ \Phi(w)f(y_2 \mid \textbf{x}) \right]^{y_1} \end{eqnarray}$

f (y_{2} ∣ z)

$f(y_2 \mid \textbf{z})$

y_{1}

$y_1$

maximum-likelihood econometrics probit

— Cederlöf
źródło

(\begin{matrix} X \\ Y \end{matrix}) \sim N ([\begin{matrix} μ_{X} \\ μ_{Y} \end{matrix}], [\begin{matrix} σ_{X}^{2} & ρ σ_{X} σ_{Y} \\ ρ σ_{X} σ_{Y} & σ_{Y}^{2} \end{matrix}]),

$\begin{pmatrix}X \\ Y\end{pmatrix}\sim\mathcal{N}\left(\begin{bmatrix}\mu_X\\\mu_Y\end{bmatrix}, \begin{bmatrix}\sigma_X^2 & \rho\sigma_X\sigma_Y\\\rho\sigma_X\sigma_Y & \sigma_Y^2\end{bmatrix}\right),$

Y

$Y$

X

$X$

Y ∣ X \sim N (μ_{Y} + ρ σ_{Y} \frac{X - μ_{X}}{σ_{X}}, σ_{Y} [1 - ρ^{2}]) .

$Y\mid X \sim \mathcal{N}\left(\mu_Y+\rho\sigma_Y\frac{X-\mu_X}{\sigma_X},\sigma_Y\left[1-\rho^2\right]\right).$

In the present case, we have

\begin{aligned} u_{1} ∣ v_{2} & \sim N (0 + \frac{η}{1 \cdot τ} \cdot 1 \frac{v_{2} - 0}{τ}, 1 \cdot [1 - {(\frac{η}{1 \cdot τ})}^{2}]) \\ = N (\frac{η}{τ^{2}} v_{2}, 1 - \frac{η^{2}}{τ^{2}}), \end{aligned}

$\begin{align} u_1 \mid v_2 &\sim \mathcal{N}\left(0+\frac{\eta}{1\cdot\tau}\cdot1\frac{v_2-0}{\tau}, 1\cdot\left[1-\left(\frac{\eta}{1\cdot\tau}\right)^2\right] \right) \\ &= \mathcal{N}\left(\frac{\eta}{\tau^2}v_2, 1-\frac{\eta^2}{\tau^2} \right), \end{align}$ which means that

u_{1} = \frac{η}{τ^{2}} v_{2} + ξ

$u_1=\frac{\eta}{\tau^2}v_2+\xi$ where (and this was your first mistake)

ξ \sim N (0, 1 - \frac{η^{2}}{τ^{2}}) .

$\xi\sim\mathcal{N}\left(0,1-\frac{\eta^2}{\tau^2}\right).$

We can thus rewrite the first equation

\begin{aligned} y_{1}^{*} & = δ_{1} z_{1} + α_{1} y_{2} + u_{1} \\ = δ_{1} z_{1} + α_{1} y_{2} + \frac{η}{τ^{2}} v_{2} + ξ \\ = δ_{1} z_{1} + α_{1} y_{2} + \frac{η}{τ^{2}} (y_{2} - z δ) + ξ . \end{aligned}

$\begin{align} y_1^* &= \delta_1 z_1 + \alpha_1 y_2 + u_1 \\ &= \delta_1 z_1 + \alpha_1 y_2 + \frac{\eta}{\tau^2}v_2+\xi \\ &= \delta_1 z_1 + \alpha_1 y_2 + \frac{\eta}{\tau^2}(y_2-\textbf{z}\delta)+\xi. \end{align}$

Now, remember that the conditional probability density function of $X=x$ given $Y=y$ is

f_{X} (x ∣ y) = \frac{f_{X Y} (x, y)}{f_{Y} (y)} .

$f_{X}(x \mid y)=\frac{f_{XY}(x,y)}{f_{Y}(y)}.$

In the present case, we have

f_{1} (y_{1} ∣ y_{2}, z) = \frac{f_{12} (y_{1}, y_{2} ∣ z)}{f_{2} (y_{2} ∣ z)},

$f_{1}(y_1 \mid y_2, \mathbf{z})=\frac{f_{12}(y_1,y_2 \mid \mathbf{z})}{f_{2}(y_2 \mid \mathbf{z})},$ which can be rearranged to your expression

f_{12} (y_{1}, y_{2} ∣ z) = f_{1} (y_{1} ∣ y_{2}, z) f_{2} (y_{2} ∣ z) .

$f_{12}(y_1, y_2 \mid \mathbf{z})= f_{1}(y_1 \mid y_2, \mathbf{z})f_{2}(y_2 \mid \mathbf{z}).$

Then, we can write the likelihood as a function of the densities of the two independent shocks $v_1,\xi_1$ :

\begin{aligned} L (y_{1}, y_{2} ∣ z) & = \prod_{i}^{n} f_{1} (y_{1 i} ∣ y_{2 i}, z_{i}) f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} Pr {(y_{1 i} = 1)}^{y_{1 i}} Pr {(y_{1 i} = 0)}^{1 - y_{1 i}} f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} Pr {(y_{1 i}^{*} > 0)}^{y_{1 i}} Pr {(y_{1 i}^{*} \leq 0)}^{1 - y_{1 i}} f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} Pr {(δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ) + ξ_{i} > 0)}^{y_{1 i}} \\ Pr {(δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ) + ξ_{i} \leq 0)}^{1 - y_{1 i}} \\ f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} Pr {(ξ_{i} > - [δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ)])}^{y_{1 i}} \\ Pr {(ξ_{i} \leq - [δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ)])}^{1 - y_{1 i}} \\ f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} Pr {(\frac{ξ_{i} - 0}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} > - \frac{δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ) + 0}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}})}^{y_{1 i}} \\ Pr {(\frac{ξ_{i} - 0}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} \leq - \frac{δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ) + 0}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}})}^{1 - y_{1 i}} \\ f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} Pr {(\frac{ξ_{i}}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} > - w_{i})}^{y_{1 i}} Pr {(\frac{ξ_{i}}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} \leq - w_{i})}^{1 - y_{1 i}} f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i}^{n} {[1 - Pr (\frac{ξ_{i}}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} \leq - w_{i})]}^{y_{1 i}} Pr {(\frac{ξ_{i}}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} \leq - w_{i})}^{1 - y_{1 i}} f_{2} (y_{2 i} ∣ z_{i}) \\ = \prod_{i} {[1 - Φ (- w_{i})]}^{y_{1 i}} Φ (- w_{i})^{1 - y_{1 i}} φ (\frac{y_{2 i} - z_{i} δ}{τ}) \\ = \prod_{i}^{n} Φ (w_{i})^{y_{1 i}} {[1 - Φ (w_{i})]}^{1 - y_{1 i}} φ (\frac{y_{2 i} - z_{i} δ}{τ}) \\ = Φ (w)^{y_{1}} {[1 - Φ (w)]}^{1 - y_{1}} φ (\frac{y_{2} - z δ}{τ}) \end{aligned}

$\begin{align} \mathcal{L}(y_1,y_2\mid \mathbf{z}) &= \prod_i^n f_{1}(y_{1i} \mid y_{2i}, \mathbf{z}_i)f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \Pr\left(y_{1i}=1\right)^{y_{1i}}\Pr\left(y_{1i}=0\right)^{1-y_{1i}}f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \Pr\left(y_{1i}^*>0\right)^{y_{1i}}\Pr\left(y_{1i}^*\leq0\right)^{1-y_{1i}}f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \Pr\left(\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_{i}\delta)+\xi_i>0\right)^{y_{1i}}\\ &\qquad\quad \Pr\left(\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_i\delta)+\xi_i\leq0\right)^{1-y_{1i}}\\ &\qquad\quad f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \Pr\left(\xi_i>-\left[\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_i\delta)\right]\right)^{y_{1i}}\\ &\qquad\quad \Pr\left(\xi_i\leq-\left[\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_i\delta)\right]\right)^{1-y_{1i}}\\ &\qquad\quad f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \Pr\left(\frac{\xi_i-0}{\sqrt{1-\frac{\eta^2}{\tau^2}}}>-\frac{\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_i\delta)+0}{\sqrt{1-\frac{\eta^2}{\tau^2}}}\right)^{y_{1i}}\\ &\qquad\quad \Pr\left(\frac{\xi_i-0}{\sqrt{1-\frac{\eta^2}{\tau^2}}}\leq-\frac{\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_i\delta)+0}{\sqrt{1-\frac{\eta^2}{\tau^2}}}\right)^{1-y_{1i}}\\ &\qquad\quad f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \Pr\left(\frac{\xi_i}{\sqrt{1-\frac{\eta^2}{\tau^2}}}>-w_i\right)^{y_{1i}} \Pr\left(\frac{\xi_i}{\sqrt{1-\frac{\eta^2}{\tau^2}}}\leq-w_i\right)^{1-y_{1i}} f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i^n \left[1-\Pr\left(\frac{\xi_i}{\sqrt{1-\frac{\eta^2}{\tau^2}}}\leq-w_i\right)\right]^{y_{1i}} \Pr\left(\frac{\xi_i}{\sqrt{1-\frac{\eta^2}{\tau^2}}}\leq-w_i\right)^{1-y_{1i}} f_{2}(y_{2i} \mid \mathbf{z}_i) \\ &= \prod_i \left[1-\Phi(-w_i)\right]^{y_{1i}} \Phi(-w_i)^{1-y_{1i}} \varphi\left(\frac{y_{2i}-\mathbf{z}_i\delta}{\tau}\right) \\ &= \prod_i^n \Phi(w_i)^{y_{1i}} \left[1-\Phi(w_i)\right]^{1-y_{1i}} \varphi\left(\frac{y_{2i}-\mathbf{z}_i\delta}{\tau}\right) \\ &= \Phi(w)^{y_{1}} \left[1-\Phi(w)\right]^{1-y_{1}} \varphi\left(\frac{y_{2}-\mathbf{z}\delta}{\tau}\right) \\ \end{align}$ where

\begin{aligned} w_{i} = \frac{δ_{1} z_{1 i} + α_{1} y_{2 i} + \frac{η}{τ^{2}} (y_{2 i} - z_{i} δ)}{\sqrt{1 - \frac{η^{2}}{τ^{2}}}} . \end{aligned}

$\begin{align} w_i = \frac{\delta_1 z_{1i} + \alpha_1 y_{2i} + \frac{\eta}{\tau^2}(y_{2i}-\textbf{z}_i\delta)}{\sqrt{1-\frac{\eta^2}{\tau^2}}}. \end{align}$

Φ (z)

$\Phi(z)$ and

φ (z)

$\varphi(z)$ are the cumulative density function and probability density function of the standard normal distribution.

— Fredrik P
źródło