19. augusta 2012
V týchto modeloch využijeme diferenciálnu transformáciu.
>model3<-lm(diff(robbery)~diff(afam)+diff(population)+diff(income)+diff(density),u)
> summary(model3)
Call:
lm(formula = diff(robbery) ~ diff(afam) + diff(population) +
diff(income) + diff(density), data = u)
Residuals:
Min 1Q Median 3Q Max
-4.623 -2.230 0.070 1.518 5.615
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 5.655e-01 1.562e+00 0.362 0.722
diff(afam) -1.238e+01 2.274e+01 -0.544 0.593
diff(population) -4.417e+03 7.694e+03 -0.574 0.573
diff(income) -2.699e-03 1.776e-03 -1.520 0.147
diff(density) 3.498e+05 5.864e+05 0.596 0.559
Residual standard error: 3.62 on 17 degrees of freedom
Multiple R-squared: 0.2085, Adjusted R-squared: 0.02222
F-statistic: 1.119 on 4 and 17 DF, p-value: 0.38
Všetky vysvetľujúce premenné sú málo významné, preto vyhadzujem afam, pretože má najvyššiu hodnotu p.
>model3.1<-lm(diff(robbery)~diff(population)+diff(income)+diff(density),u)
> summary(model3.1)
Call:
lm(formula = diff(robbery) ~ diff(population) + diff(income) +
diff(density), data = u)
Residuals:
Min 1Q Median 3Q Max
-5.2003 -2.1339 -0.0041 1.5906 5.5832
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -6.051e-02 1.037e+00 -0.058 0.9541
diff(population) -3.741e+03 7.443e+03 -0.503 0.6214
diff(income) -3.038e-03 1.631e-03 -1.863 0.0789 .
diff(density) 2.996e+05 5.677e+05 0.528 0.6041
—
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 3.549 on 18 degrees of freedom
Multiple R-squared: 0.1947, Adjusted R-squared: 0.06044
F-statistic: 1.45 on 3 and 18 DF, p-value: 0.2614
Ešte vždy sa v modeli nachádzajú málo významné premenné, vyhadzujem population , pretože má najvyššiu hodnotu p.
> model3.2<-lm(diff(robbery)~diff(income)+diff(density),u)
> summary(model3.2)
Call:
lm(formula = diff(robbery) ~ diff(income) + diff(density), data = u)
Residuals:
Min 1Q Median 3Q Max
-5.2981 -2.1578 -0.0934 1.9891 5.5808
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 4.423e-02 9.959e-01 0.044 0.9650
diff(income) -2.958e-03 1.591e-03 -1.860 0.0785 .
diff(density) 1.451e+04 2.025e+04 0.717 0.4823
—
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 3.478 on 19 degrees of freedom
Multiple R-squared: 0.1834, Adjusted R-squared: 0.0974
F-statistic: 2.133 on 2 and 19 DF, p-value: 0.1460
Ešte vždy sa v modeli nachádzajú málo významné premenné, vyhadzujem density , pretože má najvyššiu hodnotu p.
> model3.3<-lm(diff(robbery)~diff(income),u)
> summary(model3.3)
Call:
lm(formula = diff(robbery) ~ diff(income), data = u)
Residuals:
Min 1Q Median 3Q Max
-5.2171 -1.7167 -0.4599 1.7368 6.1416
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.459348 0.800248 0.574 0.572
diff(income) -0.003067 0.001564 -1.961 0.064 .
—
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 3.436 on 20 degrees of freedom
Multiple R-squared: 0.1613, Adjusted R-squared: 0.1193
F-statistic: 3.846 on 1 and 20 DF, p-value: 0.06394
V modeli 3.3 už máme významné premenné, preto tento model ideme ďalej testovať.
Testujeme heteroskedasticitu pomocou bptest :
H0: nieje heteroskedasticita
H1: je heteroskedasticita
> bptest(model3.3)
studentized Breusch-Pagan test
data: model3.3
BP = 2.0019, df = 1, p-value = 0.1571
P hodnota je väčšia ako α, H0 nezamietame. Môžeme predpokladať, že v modeli 3.3 nieje heteroskedasticita.
Testujeme autokoreláciu pomocou dwtest :
H0: nieje autokorelácia
H1: je autokorelácia
> dwtest(model3.3,alternative=”two.sided”)
Durbin-Watson test
data: model3.3
DW = 2.1541, p-value = 0.6741
alternative hypothesis: true autocorelation is not 0
DW hodnota je blízko 2,H0 nezamietame, v modeli nie je autokorelacia.
Nieje možné testovať multikolinearitu, pretože máme len jednu premennú.
Záver: Model 3.3 zamietame.
Musíte byť prihlásený, aby ste mohli zverejniť komentár.