-
Notifications
You must be signed in to change notification settings - Fork 0
/
hw4p1-why-people-vote.Rmd
145 lines (115 loc) · 2.81 KB
/
hw4p1-why-people-vote.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
---
title: "Why People Vote"
author: "Terrel Shumway"
date: "05/02/2015"
output: html_document
---
This document presents answers for homework 4 part 1.
## Getting and Cleaning the Data
```{r}
library(data.table)
baseurl = "https://courses.edx.org/c4x/MITx/15.071x_2/asset/"
getdata = function(local){
if(!file.exists(local)){
library(downloader)
remote = paste0(baseurl,local)
print(remote)
download(remote,local)
}
read.csv(local)
}
data = getdata("gerber.csv")
# this is for my sanity
data$sex = factor(data$sex)
levels(data$sex) = c("male","female")
```
## Preparing the Dataset
problem 1.1:
```{r}
depvar="voting"
idvars = c("sex","yob")
mean(data[,depvar])
```
problem 1.2:
```{r}
# first, convert from indicator matrix to factor
# thanks to http://stackoverflow.com/users/684229/tms
# for this clever code
# http://stackoverflow.com/questions/7727315/how-to-create-a-factor-from-a-binary-indicator-matrix
groups = c("control","civicduty","hawthorne","self","neighbors")
data$stimulus = as.factor(groups[as.matrix(data[,groups]) %*% 1:length(groups)])
# get the mean of each group
sort(tapply(data$voting,data$stimulus,mean))
```
problem 1.3:
```{r}
m1 = glm(voting~stimulus,data=data,family="binomial")
summary(m1)
```
problem 1.4:
```{r}
m1.pred = predict(m1,type="response")
conf = table(data[,depvar],m1.pred>0.3)
sum(diag(conf))/nrow(data)
```
problem 1.5:
```{r}
conf = table(data[,depvar],m1.pred>0.5)
sum(diag(conf))/nrow(data)
```
problem 1.6
```{r}
library(ROCR)
pred = prediction(m1.pred,data[,depvar])
plot(performance(pred,"tpr","fpr"))
as.numeric(performance(pred,"auc")@y.values)
```
## Trees
```{r}
library(rpart)
library(rpart.plot)
```
problem 2.1
```{r}
m2 = rpart(voting~stimulus,data=data)
prp(m2)
```
```{r}
m3 = rpart(voting~stimulus,data=data,cp=0)
prp(m3)
```
problem 2.2,2.3:
```{r}
# this model is identical to `m3` above, just a different shape
m4 = rpart(voting~civicduty+hawthorne+self+neighbors,data=data,cp=0)
prp(m4)
```
problem 2.4:
```{r}
m5 = rpart(voting~stimulus+sex,data=data,cp=0)
prp(m5)
```
```{r}
# this model is identical to `m3` above, just a different shape
m6 = rpart(voting~sex+civicduty+hawthorne+self+neighbors,data=data,cp=0)
prp(m6)
```
## Interaction Terms
```{r}
P = data.frame(sex=c(0,0,1,1),control=c(0,1,0,1))
P$sex = factor(P$sex)
levels(P$sex) = c("male","female")
m7 = rpart(voting ~ control,data=data,cp=0)
#prp(m7,digits=6)
cbind(P,pred=predict(m7,newdata=P,method="response"))
m8 = rpart(voting ~ control+sex,data=data,cp=0)
#prp(m8,digits=6)
cbind(P,pred=predict(m8,newdata=P,method="response"))
```
```{r}
m9 = glm(voting~control+sex,data=data,family="binomial")
cbind(P,pred=predict(m9,newdata=P,type="response"))
m10 = glm(voting~control+sex+sex:control,data=data,family="binomial")
cbind(P,pred=predict(m10,newdata=P,type="response"))
summary(m10)
```