Wzory 2
Macierzowa postać modelu
$$\begin{bmatrix}
1 & 0 \\
{- \beta}_{21} & 1 \\
\end{bmatrix}\begin{bmatrix}
Y_{1} \\
Y_{2} \\
\end{bmatrix} = \begin{bmatrix}
\gamma_{11} & 0 & \gamma_{13} \\
0 & \gamma_{22} & 0 \\
\end{bmatrix}\begin{bmatrix}
Z_{1} \\
Z_{2} \\
Z_{3} \\
\end{bmatrix} + \begin{bmatrix}
\varepsilon_{1} \\
\varepsilon_{1} \\
\end{bmatrix}$$
β Γ
Parametry strukturalne
Równanie zrewidowane:
$$Y_{2} = \beta_{21}{\hat{Y}}_{1} + \gamma_{21}Z_{1} + \gamma_{23}Z_{3} + \varepsilon_{2}$$
$c_{2} = \begin{bmatrix} \begin{matrix} b_{21} \\ c_{21} \\ \end{matrix} \\ c_{22} \\ \end{bmatrix}$ wektor estymatorów wektora parametrów strukturalnych $\gamma = \begin{bmatrix} \begin{matrix} \beta_{21} \\ \gamma_{21} \\ \end{matrix} \\ \gamma_{23} \\ \end{bmatrix}$
Wektor estymatorów:
cj = (AjTAj)-1 AjTyj
$$A_{j} = \left\lbrack {\hat{y}}_{i}\ Z_{j} \right\rbrack$$
$$\left( {A_{j}}^{T}A_{j} \right)^{- 1} = \ \frac{1}{\left| {A_{j}}^{T}A_{j} \right|}\left( {A_{j}}^{T}A_{j} \right)^{D}$$
Postać teoretyczna:
$${\hat{Y}}_{2} = b_{21}{\hat{Y}}_{1} + c_{21}Z_{1} + c_{23}Z_{3}$$
Estymator wariancji składnika losowego: $S^{2}\left( e_{2} \right) = \frac{1}{n - k_{2}} \bullet {(y_{2}}^{T}y_{2} - \ {c_{2}}^{T}{A_{2}}^{T}y_{2})$
$${y_{2}}^{T}y_{2} = \ \left\lbrack \sum_{t = 1}^{n}{y_{t2}}^{2} \right\rbrack$$
Błąd standardowy składnika losowego: $S\left( e \right) = \sqrt{S^{2}(e)}$
Macierz ocen wariancji i kowariancji estymatorów: S(c2,c2) = S2(e2)(A2TA2)−1
${\hat{Y}}_{2} = \ \frac{b_{21}{\hat{Y}}_{1}}{\left( S\left( b_{21} \right) \right)} + \ \frac{c_{21}Z_{1}}{\left( S\left( c_{21} \right) \right)} + \frac{c_{23}Z_{3}}{\left( S\left( c_{23} \right) \right)}\text{\ \ \ }\frac{\ }{\left( S\left( e_{2} \right) \right)}$
Miary unormowania dopasowania modelu do danych empirycznych:
$\left\{ \begin{matrix} {\Phi_{2}}^{2} = \frac{{y_{2}}^{T}y_{2} - \ {c_{2}}^{T}{A_{2}}^{T}y_{2}}{{y_{2}}^{T}y_{2} - n{{\overset{\overline{}}{y}}_{2}}^{2}}\text{\ \ \ \ \ } - \text{wsp.zbie}z\text{ny} \\ {R_{2}}^{2} = 1 - {\Phi_{2}}^{2}\ - \text{wsp.determinacji} \\ \end{matrix} \right.\ $
${\overset{\overline{}}{y}}_{2} = \ \frac{1}{n}\ \sum_{t = 1}^{n}y_{t2}$
Prognoza punktowa:
${\hat{y}}_{\tau 2} = b_{21}{\hat{y}}_{\tau 1} + c_{21}Z_{\tau 1} + c_{23}Z_{\tau 3}$
Standardowy błąd prognozy: $S_{\tau} = S\left( e \right)\sqrt{1 + {\tilde{z}}_{\text{τ\ }}{({A_{2}}^{T}A_{2})}^{- 1}{{\tilde{z}}_{\tau}}^{T}}$
${\tilde{z}}_{\text{τ\ }} = \begin{bmatrix} {\hat{y}}_{\tau 1} & z_{\tau 1} & z_{\tau 3} \\ \end{bmatrix}$
Względny błąd prognozy: $\ V_{\tau 2} = \frac{S_{\tau}}{|{\hat{y}}_{\tau}|} \bullet 100\%$