The k-fold cross-validation is implemented by taking a certain number of observations defined by n and randomly splitting them into groups defined by K. The groups that are split act as a validation set and the remainder act as a validation set. The test error of the data is then calculated by averaging the k resulting MSE estimates.
1.The validation set approach
There are two drawbacks to using the variable set approach, the first is that the estimate of the test error could be highly variablke based on the data that is inputed into the training and vailidation data sets. Secondly, the vailidation set error rate may tend to overestimate the test error rate for the model fit of the entire data set.
The LOOCV approach has an advatage as it will produce less bias. The validation approach would produce diffrent MSE when applied multiple times due to the splitting of the data, while performing LOOCV this process will yield the same results. Due to the fact that we are splitting the data based on 1 observation. Another disadvantage this model has is that is computationally intensive.
library(ISLR)
attach(Default)
set.seed(1)
fit.glm = glm(default ~ income + balance, data = Default, family = "binomial")
summary(fit.glm)
##
## Call:
## glm(formula = default ~ income + balance, family = "binomial",
## data = Default)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.4725 -0.1444 -0.0574 -0.0211 3.7245
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.154e+01 4.348e-01 -26.545 < 2e-16 ***
## income 2.081e-05 4.985e-06 4.174 2.99e-05 ***
## balance 5.647e-03 2.274e-04 24.836 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2920.6 on 9999 degrees of freedom
## Residual deviance: 1579.0 on 9997 degrees of freedom
## AIC: 1585
##
## Number of Fisher Scoring iterations: 8
train = sample(dim(Default)[1], dim(Default)[1]/2)
2.Fit a multiple logistic regression model using only the training observations.
fit.glm = glm(default ~ income + balance, data = Default, family = "binomial", subset = train)
summary(fit.glm)
##
## Call:
## glm(formula = default ~ income + balance, family = "binomial",
## data = Default, subset = train)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.5830 -0.1428 -0.0573 -0.0213 3.3395
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.194e+01 6.178e-01 -19.333 < 2e-16 ***
## income 3.262e-05 7.024e-06 4.644 3.41e-06 ***
## balance 5.689e-03 3.158e-04 18.014 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 1523.8 on 4999 degrees of freedom
## Residual deviance: 803.3 on 4997 degrees of freedom
## AIC: 809.3
##
## Number of Fisher Scoring iterations: 8
3.Obtain a prediction of default status for each individual in the validation set by computing the posterior probability of default for that individual, and classifying the individual to the default category if the posterior probability is greater than 0.5.
probs = predict(fit.glm, newdata = Default[-train,], type = "response")
pred.glm = rep("No", length(probs))
pred.glm[probs > .5] = "yes"
4.Compute the validation set error, which is the fraction of the observations in the validation set that are misclassified.
mean(pred.glm != Default[-train,]$default)
## [1] 0.0352
The validation set approach gave me a test rate error if 3.52%
train = sample(dim(Default)[1], dim(Default)[1]/2)
fit.glm = glm(default ~ income + balance, data = Default, family = "binomial", subset = train)
probs = predict(fit.glm, newdata = Default[-train,], type = "response")
pred.glm = rep("No", length(probs))
pred.glm[probs > .5] = "yes"
mean(pred.glm != Default[-train,]$default)
## [1] 0.0374
train = sample(dim(Default)[1], dim(Default)[1]/2)
fit.glm = glm(default ~ income + balance, data = Default, family = "binomial", subset = train)
probs = predict(fit.glm, newdata = Default[-train,], type = "response")
pred.glm = rep("No", length(probs))
pred.glm[probs > .5] = "yes"
mean(pred.glm != Default[-train,]$default)
## [1] 0.0346
train = sample(dim(Default)[1], dim(Default)[1]/2)
fit.glm = glm(default ~ income + balance, data = Default, family = "binomial", subset = train)
probs = predict(fit.glm, newdata = Default[-train,], type = "response")
pred.glm = rep("No", length(probs))
pred.glm[probs > .5] = "yes"
mean(pred.glm != Default[-train,]$default)
## [1] 0.0352
These three diffrent splits produced a diffrent resulting error rate which shows that the rate does in fact vary based on which observations are in the training/validation sets.
train = sample(dim(Default)[1], dim(Default)[1]/2)
fit.glm = glm(default ~ income + balance + student, data = Default, family = "binomial", subset = train)
probs = predict(fit.glm, newdata = Default[-train,], type = "response")
pred.glm = rep("No", length(probs))
pred.glm[probs > .5] = "yes"
mean(pred.glm != Default[-train,]$default)
## [1] 0.0384
When including the dummy variable student it does not seem to lower the test error rate.
set.seed(1)
attach(Default)
## The following objects are masked from Default (pos = 3):
##
## balance, default, income, student
fit.glm = glm(default ~ income + balance, data = Default, family = "binomial")
summary(fit.glm)
##
## Call:
## glm(formula = default ~ income + balance, family = "binomial",
## data = Default)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.4725 -0.1444 -0.0574 -0.0211 3.7245
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.154e+01 4.348e-01 -26.545 < 2e-16 ***
## income 2.081e-05 4.985e-06 4.174 2.99e-05 ***
## balance 5.647e-03 2.274e-04 24.836 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 2920.6 on 9999 degrees of freedom
## Residual deviance: 1579.0 on 9997 degrees of freedom
## AIC: 1585
##
## Number of Fisher Scoring iterations: 8
The standard error for the income coeffiecent is 4.985e-06 and the standard error for the balance coeffiecent was 2.274e-04.
boot.fn = function(data, index) {
fit = glm(default ~ income + balance, data = data, family = "binomial", subset = index)
return(coef(fit))
}
library(boot)
boot(Default, boot.fn, 1000)
##
## ORDINARY NONPARAMETRIC BOOTSTRAP
##
##
## Call:
## boot(data = Default, statistic = boot.fn, R = 1000)
##
##
## Bootstrap Statistics :
## original bias std. error
## t1* -1.154047e+01 -3.945460e-02 4.344722e-01
## t2* 2.080898e-05 1.680317e-07 4.866284e-06
## t3* 5.647103e-03 1.855765e-05 2.298949e-04
The standard errors for the coefficients are 4.866284e-06 and 2.298949e-04.
The standard error for both functions are relatively similar.
library(MASS)
attach(Boston)
sample_mu = mean(medv)
sample_mu
## [1] 22.53281
This displays the sample mean of 22.53281
sample_se = sd(medv)/sqrt(nrow(Boston))
sample_se
## [1] 0.4088611
The standard error estimate is .4088611. This standard error will give us the accuracy of the estimate how much the mean will vary if a different sample was chosen.
library(boot)
set.seed(1)
boot.fn = function(data,index){
mu = mean(data[index])
return(mu)
}
set.seed(1)
boot(medv,boot.fn, 100)
##
## ORDINARY NONPARAMETRIC BOOTSTRAP
##
##
## Call:
## boot(data = medv, statistic = boot.fn, R = 100)
##
##
## Bootstrap Statistics :
## original bias std. error
## t1* 22.53281 0.009027668 0.3482331
The standard error obtained from bootstrapping is slightly less than the one obtained above.
t.test(medv)
##
## One Sample t-test
##
## data: medv
## t = 55.111, df = 505, p-value < 2.2e-16
## alternative hypothesis: true mean is not equal to 0
## 95 percent confidence interval:
## 21.72953 23.33608
## sample estimates:
## mean of x
## 22.53281
ci.mu = c(sample_mu-2*0.3482331,sample_mu+2*0.3482331)
ci.mu
## [1] 21.83634 23.22927
The confidence interval obtained is very similar to the one from the t-test.
mu_median = median(medv)
mu_median
## [1] 21.2
The estimate median is 21.2
library(boot)
set.seed(1)
boot.fn = function(data,index){
mu.median = median(data[index])
return(mu.median)
}
set.seed(1)
boot(medv,boot.fn, 100)
##
## ORDINARY NONPARAMETRIC BOOTSTRAP
##
##
## Call:
## boot(data = medv, statistic = boot.fn, R = 100)
##
##
## Bootstrap Statistics :
## original bias std. error
## t1* 21.2 -0.029 0.3461316
Here we computed the standard error of the median using the boostrapping method.
mu.quantile = quantile(medv,c(.1))
mu.quantile
## 10%
## 12.75
The estimate for the tenth percentile of medv in boston suburbs is 12.75.
boot.fn = function(data,index){
mu.quantile = quantile(data[index], c(.1))
return(mu.quantile)
}
boot(medv, boot.fn,1000)
##
## ORDINARY NONPARAMETRIC BOOTSTRAP
##
##
## Call:
## boot(data = medv, statistic = boot.fn, R = 1000)
##
##
## Bootstrap Statistics :
## original bias std. error
## t1* 12.75 0.0285 0.4861472
This is showing the standard error of the tenth percentile using the bootstrapping method. I observed that the standard error was 0.4917843 which is very low telling us that this method is very accurate.