diff --git a/hw3/.RData b/hw3/.RData
index 75b0d6a..97b22f9 100644
Binary files a/hw3/.RData and b/hw3/.RData differ
diff --git a/hw3/.Rhistory b/hw3/.Rhistory
index 3a641a7..ee5aad7 100644
--- a/hw3/.Rhistory
+++ b/hw3/.Rhistory
@@ -77,3 +77,81 @@ length(test_data)
 ncol(test_data)
 nrow(test_data)
 q()
+qda.fit
+fit.qda
+qda.fit
+auto
+qda.fit = qda(mpg01 ~ horsepower + weight + cylinders + displacement, data = train_data)
+import(MASS)
+qda.fit = qda(mpg01 ~ horsepower + weight + cylinders + displacement, data = train_data)
+import(class)
+library(MASS)
+qda.fit = qda(mpg01 ~ horsepower + weight + cylinders + displacement, data = train_data)
+qda.fit
+> mean(qda.pred$class!=test_data$mpg01,na.rm=T)
+mean(qda.pred$class!=test_data$mpg01,na.rm=T)
+qda.fit=qda(mpg01 ~ horsepower + weight + acceleration + displacement,data=train_data)
+qda.pred=predict(qda.fit,test_data,na.rm=T)
+mean(qda.pred$class!=test_data$mpg01,na.rm=T)
+qda.fit=qda(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data)
+qda.pred=predict(qda.fit,test_data,na.rm=T)
+mean(qda.pred$class!=test_data$mpg01,na.rm=T)
+qda.fit
+qda.pred=predict(qda.fit,test_data,na.rm=T)
+mean(qda.pred$class!=test_data$mpg01,na.rm=T)
+glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
+glm.probs=predict(glm.fit,test_data,type="response")
+glm.pred=rep(0,199)
+glm.pred[glm.probs>.5]=1
+mean(glm.pred!=test_data$mpg01)
+glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
+glm.probs=predict(glm.fit,test_data,type="response")
+glm.pred=rep(0,length(test_data)
+glm.pred[glm.probs>.5]=1
+mean(glm.pred!=test_data$mpg01)
+glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
+glm.probs=predict(glm.fit,test_data,type="response")
+glm.pred=rep(0,length(test_data))
+glm.pred[glm.probs>.5]=1
+mean(glm.pred!=test_data$mpg01)
+glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
+glm.probs=predict(glm.fit,test_data,type="response")
+glm.pred=rep(0,length(test_data))
+glm.pred[glm.probs>.5]=1
+mean(glm.pred!=test_data$mpg01)
+glm.pred
+glm.pred=rep(0,length(test_data))
+glm.pred
+test_data
+glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
+glm.probs=predict(glm.fit,test_data,type="response")
+glm.pred=rep(0,nrow(test_data))
+glm.pred[glm.probs>.5]=1
+mean(glm.pred!=test_data$mpg01)
+set.seed(1)
+auto <- na.omit(auto)
+train_bools <- (auto$year %% 2 == 0)
+train_data = auto[train_bools,]
+test_data = auto[!train_bools,]
+train.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$cylinders)[train_bools,]
+test.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$cylinders)[!train_bools,]
+train.mpg01 = auto$mpg01[train_bools]
+knn.pred = knn(train.X,test.X,train.mpg01,k=1)
+mean(knn.pred != test_data$mpg01)
+knn.pred = knn(train.X,test.X,train.mpg01,k=2)
+mean(knn.pred != test_data$mpg01)
+knn.pred = knn(train.X,test.X,train.mpg01,k=3)
+mean(knn.pred != test_data$mpg01)
+knn.pred = knn(train.X,test.X,train.mpg01,k=4)
+mean(knn.pred != test_data$mpg0)
+import(class)
+library(class)
+knn.pred = knn(train.X,test.X,train.mpg01,k=1)
+mean(knn.pred != test_data$mpg01)
+knn.pred = knn(train.X,test.X,train.mpg01,k=2)
+mean(knn.pred != test_data$mpg01)
+knn.pred = knn(train.X,test.X,train.mpg01,k=3)
+mean(knn.pred != test_data$mpg01)
+knn.pred = knn(train.X,test.X,train.mpg01,k=4)
+mean(knn.pred != test_data$mpg0)
+q()
diff --git a/hw3/answers b/hw3/answers
index a4e2011..56b2827 100644
--- a/hw3/answers
+++ b/hw3/answers
@@ -196,7 +196,7 @@ Part B: Choose one of Questions 10 or 11
 
         > qda.fit
         Call:
-        lda(mpg01 ~ horsepower + weight + cylinders + displacement, data = train_data)
+        qda(mpg01 ~ horsepower + weight + cylinders + displacement, data = train_data)
         
         Prior probabilities of groups:
                 0         1 
@@ -206,17 +206,12 @@ Part B: Choose one of Questions 10 or 11
           horsepower   weight cylinders displacement
         0  131.96939 3579.827  6.755102     268.4082
         1   77.96429 2313.598  4.071429     111.7188
-        
-        Coefficients of linear discriminants:
-                               LD1
-        horsepower    0.0060634365
-        weight       -0.0011442212
-        cylinders    -0.6390942259
-        displacement  0.0004517291
+
+        > qda.pred=predict(qda.fit,test_data,na.rm=T)
 
     ***Test Data Error Rate:
         > mean(qda.pred$class!=test_data$mpg01,na.rm=T)
-        [1] 0.1428571
+        [1] 0.1483516
 
 
 
@@ -227,12 +222,12 @@ Part B: Choose one of Questions 10 or 11
 
         > glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
         > glm.probs=predict(glm.fit,test_data,type="response")
-        > glm.pred=rep(0,199)
+        > glm.pred=rep(0,nrow(test_data))
         > glm.pred[glm.probs>.5]=1
 
      ***Test Data Error Rate:
         > mean(glm.pred!=test_data$mpg01)
-        [1] 0.1407035
+        [1] 0.1373626
 
 
 ──────────────────────────────────────────────────────────────────────────
@@ -249,24 +244,30 @@ Part B: Choose one of Questions 10 or 11
         > train_bools <- (auto$year %% 2 == 0)
         > train_data = auto[train_bools,]
         > test_data = auto[!train_bools,]
-
-        > train.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$acceleration)[train_bools,]
-        > test.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$acceleration)[!train_bools,]
+        
+        > train.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$cylinders)[train_bools,]
+        > test.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$cylinders)[!train_bools,]
         > train.mpg01 = auto$mpg01[train_bools]
 
+
      ***Test Data Error Rates:
      k = 1
+        > knn.pred = knn(train.X,test.X,train.mpg01,k=1)
         > mean(knn.pred != test_data$mpg01)
         [1] 0.1483516
      k = 2
+        > knn.pred = knn(train.X,test.X,train.mpg01,k=2)
         > mean(knn.pred != test_data$mpg01)
         [1] 0.1593407
      k = 3
+        > knn.pred = knn(train.X,test.X,train.mpg01,k=3)
         > mean(knn.pred != test_data$mpg01)
         [1] 0.1648352
      k = 4
+        > knn.pred = knn(train.X,test.X,train.mpg01,k=4)
         > mean(knn.pred != test_data$mpg0)
-        [1] 0.1813187
+        [1] 0.1923077
+
 
         k = 1 looks like the best, since the error rate increases with k.
 
diff --git a/hw3/commands b/hw3/commands
index 6991c04..9505819 100644
--- a/hw3/commands
+++ b/hw3/commands
@@ -22,13 +22,13 @@ lda.fit=lda(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_da
 lda.pred=predict(lda.fit, test_data)
 mean(lda.pred$class!=test_data$mpg01,na.rm=T)
 
-qda.fit=qda(mpg01 ~ horsepower + weight + acceleration + displacement,data=train_data)
+qda.fit=qda(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data)
 qda.pred=predict(qda.fit,test_data,na.rm=T)
 mean(qda.pred$class!=test_data$mpg01,na.rm=T)
 
-glm.fit=glm(mpg01 ~ horsepower + weight + acceleration + displacement,data=train_data,family=binomial)
+glm.fit=glm(mpg01 ~ horsepower + weight + cylinders + displacement,data=train_data,family=binomial)
 glm.probs=predict(glm.fit,test_data,type="response")
-glm.pred=rep(0,199)
+glm.pred=rep(0,nrow(test_data))
 glm.pred[glm.probs>.5]=1
 mean(glm.pred!=test_data$mpg01)
 
@@ -39,8 +39,18 @@ train_data = auto[train_bools,]
 test_data = auto[!train_bools,]
 
 
-train.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$acceleration)[train_bools,]
-test.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$acceleration)[!train_bools,]
+train.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$cylinders)[train_bools,]
+test.X = cbind(auto$horsepower,auto$displacement,auto$weight,auto$cylinders)[!train_bools,]
 train.mpg01 = auto$mpg01[train_bools]
 
 knn.pred = knn(train.X,test.X,train.mpg01,k=1)
+mean(knn.pred != test_data$mpg01)
+
+knn.pred = knn(train.X,test.X,train.mpg01,k=2)
+mean(knn.pred != test_data$mpg01)
+
+knn.pred = knn(train.X,test.X,train.mpg01,k=3)
+mean(knn.pred != test_data$mpg01)
+
+knn.pred = knn(train.X,test.X,train.mpg01,k=4)
+mean(knn.pred != test_data$mpg0)