- beleške
- Linear models with R, J.J.Faraway
- Regression analyses: theory, methods, and aplications, A. Sen, M. Srivastava
Obaveze
- seminarski
- ispit
Obaveze
-Ova funkcija minimizira \(E(Y-f(X))^2.\)
\[\begin{align*} a&=\frac{EXY-EXEY}{DX}\\ b&=EY-aEX \end{align*} \] -Metodom zamene dobijamo: \[\begin{align*} \hat{a}&=\frac{\sum X_iY_i-n\bar{X}\bar{Y}}{{\bar{S}_X^2}}=\hat{\rho}\frac{\bar{S}_X}{\bar{S}_Y}\\ \hat{b}&=\bar{Y}-\hat{a}\bar{X}. \end{align*}\]
\(X\) posmatramo kao neslučajnu veličinu i zapravo slučajnost Y potiče od odgovarajućeg šuma, odnosno \(Y=f(X)+\varepsilon\).
Pretpostavljamo
da je \(Y=aX+b+\varepsilon\);
nepoznate parametre ocenjujemo metodom najmanjih kvadrata odnosno \[\min\sum_{i=1}^n(y_i-ax_i-b)^2\]
## speed dist ## 1 4 2 ## 2 4 10 ## 3 7 4 ## 4 7 22 ## 5 8 16 ## 6 9 10 ## 7 10 18 ## 8 10 26 ## 9 10 34 ## 10 11 17
\[\hat{a}=3.93\;\; \hat{b}=-17.59 \]
-nepristrasnost \[E(\hat{a})=a\;\;E(\hat{b})=b\] -postojanost
\[\begin{align*} D(\hat{a})&=\frac{\sigma^2}{\sum_{i=1}^n(x_i-\bar{x})^2}\\ D(\hat{b})&=\frac{\sigma^2}{n}\Big(1+\frac{\bar{x}}{S^2_x}\Big)\end{align*}\]
\[e_i=y_i-\hat{y}_i\] \[\begin{align*} SSE&=\sum_{i=1}^ne^2_i\\ SSR&=\sum_{i=1}^n(\hat{y}_i-\bar{y})^2\\ SSTO&=\sum_{i=1}^n(y_i-\bar{y})^2 \end{align*}\] \[SSTO=SSE+SSR\] \[\color{red}{R^2=1-\frac{SSE}{SSTO}}\]
\[E((n-2)SSE)=\sigma^2\] \[\hat{\sigma}^2=\frac{SSE}{n-2}\]
\(\hat{\sigma}=15.38\)
Pretpostavljamo da je \(\{\varepsilon_i\}\) niz nekorelisanih i jednako raspodeljenih slučajnih veličina sa normalnom \(\mathcal{N}(0,\sigma^2)\). Tada:
-pravljenje intervala poverenja i predviđanja
## ## Call: ## lm(formula = dist ~ speed, data = cars) ## ## Residuals: ## Min 1Q Median 3Q Max ## -29.069 -9.525 -2.272 9.215 43.201 ## ## Coefficients: ## Estimate Std. Error t value Pr(>|t|) ## (Intercept) -17.5791 6.7584 -2.601 0.0123 * ## speed 3.9324 0.4155 9.464 1.49e-12 *** ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## Residual standard error: 15.38 on 48 degrees of freedom ## Multiple R-squared: 0.6511, Adjusted R-squared: 0.6438 ## F-statistic: 89.57 on 1 and 48 DF, p-value: 1.49e-12