(1) Load data

(2) attach(data-sample)

(3) summary(data-sample)

(4) ced.del <- cbind(sDel, sNoDel)

(5) summary(ced.del)

(6) duckie <- glm(ced.del ~ cat + follows + factor(class), family=binomial)

(7) duckie

(8) summary(duckie)

(9) anova(duckie, test="Chisq")

(10) plot(duckie)

Revised:

fico <- read.table("/home/brent/Documents/fico.csv", header=TRUE, sep=",",

na.strings="NA", dec=".", strip.white=TRUE)

attach(fico)

duckie <- glm(approved ~ creditScore, family=binomial)

summary(duckie)

------

> summary(duckie)

Call:

glm(formula = approved ~ creditScore, family = binomial)

Deviance Residuals:

Min 1Q Median 3Q Max

-1.408 -1.338 0.959 1.010 1.149

Coefficients:

Estimate Std. Error z value Pr(>|z|)

(Intercept) -6.223592 17.177351 -0.362 0.717

creditScore 0.009605 0.024893 0.386 0.700

(Dispersion parameter for binomial family taken to be 1)

Null deviance: 20.190 on 14 degrees of freedom

Residual deviance: 20.038 on 13 degrees of freedom

AIC: 24.038

Number of Fisher Scoring iterations: 4

Use the intercept and the credit score to solve for the predicted probability:

$\widehat{p}=\frac{{e}^{\mathrm{\mathrm{\underset{o}{\beta}}+\mathrm{x1\underset{1}{\beta}}}}}{1+{e}^{\mathrm{\mathrm{\underset{o}{\beta}}+\mathrm{x1\underset{1}{\beta}}}}}$

With the coefficient estimates we have:

$\widehat{p}=\frac{{e}^{\mathrm{\mathrm{-6.223592}+\mathrm{0.009605\underset{}{x1}}}}}{1+{e}^{\mathrm{\mathrm{-6.223592}+\mathrm{0.009605\underset{}{x1}}}}}$

## No comments:

## Post a Comment