-
Notifications
You must be signed in to change notification settings - Fork 10
/
Ch09.jl
227 lines (205 loc) · 7.07 KB
/
Ch09.jl
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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
# Julia code for Chapter 9
# Chapter 9.1 Confidence Interval
####################################################################
## Histogram of the sample average
using Distributions, Plots, Printf
gmm = MixtureModel(Normal[Normal(), Normal(5., 1.),], [0.3, 0.7])
x = range(-5., 10., length=1000)
# the population distribution
plot(x, pdf.(gmm, x), linewidth=3, label=false, title="Population distribution")
## histograms of 4 random samples of size 50
plot_array = []
for i=1:4
bin = -5:10
Y = rand(gmm, 50)
tt = @sprintf("Mean = %3.2f", mean(Y))
push!(plot_array, histogram(Y, bins=bin, label = tt, legend=:topleft))
end
plot(plot_array...)
## histogram of sample average
M = zeros(10000)
bin = range(2., 5., step=0.1)
for i = 1:10000
M[i] = mean(rand(gmm, 50))
end
histogram(M, bins=bin, label=false, title="Histogram of the sample average")
####################################################################
## Compute confidence interval
# Julia code to compute the width of the confidence interval
using Distributions
α = 0.05
μ = 0.
σ = 1.
ϵ = quantile(Normal(μ, σ ), 1. - α/2.) # CI is estimator ± ϵ
####################################################################
## Visualize the t-distribution
# Julia code to plot the t-distribution
using Distributions, Plots
x = range(-5.,5.,length=100)
p1 = pdf.(Normal(),x)
p2 = pdf.(TDist(11-1),x)
p3 = pdf.(TDist(3-1),x)
p4 = pdf.(TDist(2-1),x)
plot([p1,p2,p3,p4], label = ["Gaussian(0,1)" "t-dist, N = 11" "t-dist, N = 3" "t-dist, N = 2"], legend=:topright)
####################################################################
## Construct a confidence interval from data
# Julia code to generate a confidence interval
using Statistics, Distributions, Printf
x = [72., 69., 75., 58., 67., 70., 60., 71., 59., 65.]
θhat = mean(x) # Sample mean
σhat = std(x) # Sample standard deviation
N = size(x,1)
ν = N-1 # degrees of freedom
α = 0.05 # confidence level
z = quantile(TDist(ν), 1. - α/2.)
CI_L = θhat - z*σhat/sqrt(N)
CI_U = θhat + z*σhat/sqrt(N)
@printf("Confidence interval: [%5.3f - %5.3f]", CI_L, CI_U)
####################################################################
# Chapter 9.2 Bootstrap
####################################################################
## Visualize bootstrap
# Julia code to illustrate bootstrap
using Plots, Distributions, Statistics
gmm = MixtureModel(Normal[Normal(), Normal(5., 1.),], [0.3, 0.7])
## visualize the population
x = range(-5., 10., length=1000);
f = pdf.(gmm, x);
plot(x, pdf.(gmm,x), linewidth=3, label="Population", legend=:topleft)
## Visualize bootstrap dataset, the real sample data
N = 100
X = rand(gmm, N) # the real sample data
p1 = histogram(X, bins = range(-5,10,length=30), label=false)
## visualize 3 bootstrap samples
plot_array = []
for i=1:3
# Bootstrap dataset
idx = rand(1:N, N)
Xb1 = X[idx]
push!(plot_array, histogram(Xb1, bins = range(-5.,10.,length=30)))
end
plot(plot_array..., label=false)
# Bootstrap the median
K = 1000 # number of bootstrap re-samples
θhat = zeros(K)
for i=1:K # repeat K times
idx = rand(1:N, N) # sampling w/ replacement
Y = X[idx]
θhat[i] = median(Y) # estimator
end
M = mean(θhat) # bootstrap mean
V = var(θhat) # bootstrap variance
## histogram of bootstrap replications of estimator
histogram(θhat, bins=range(3.5,5.5,length=30), legend=false)
####################################################################
## Bootstrap median using real data
# Julia code to estimate a bootstrap variance
using Statistics
X = [72., 69., 75., 58., 67., 70., 60., 71., 59., 65.]
N = size(X,1)
K = 1000
θhat = zeros(K)
for i=1:K # repeat K times
idx = rand(1:N, N) # sampling w/ replacement
Y = X[idx]
θhat[i] = median(Y) # estimator
end
M = mean(θhat) # bootstrap mean
V = var(θhat) # bootstrap variance
####################################################################
# Chapter 9.3 Hypothesis Testing
####################################################################
## Estimate Z-value
θhat = 0.29 # Your estimate
θ = 0.25 # Your hypothesis
σ = sqrt(θ*(1. - θ)) # Known standard deviation
N = 1000 # Number of samples
Z_hat = (θhat - θ)/(σ/sqrt(N))
####################################################################
## Compute critical value
using Distributions
α = 0.05
z_α = quantile(Normal(), 1-α)
####################################################################
## Compute p-value
using Distributions
p = cdf(Normal(), -1.92)
####################################################################
# Chapter 9.5 ROC Curve
####################################################################
## Plot an ROC curve
# Julia code to plot ROC curve
using Distributions, Plots, Statistics
σ = 2.
μ = 3.
PF1 = zeros(1000); PD1 = zeros(1000)
PF2 = zeros(1000); PD2 = zeros(1000)
αset = range(0.,1.,length=1000)
d = Normal()
for i=1:1000
α = αset[i]
PF1[i] = α
PD1[i] = α
PF2[i] = α
PD2[i] = 1. - cdf(d, quantile(d, 1-α)-μ/σ)
end
plot(PF1, PD1, linewidth=3, label = "Blind guess")
plot!(PF2, PD2, linewidth=3, label = "Neyman-Pearson", legend=:bottomright)
# Computer area under curve
auc1 = sum(PD1.*[0.; diff(PF1)])
auc2 = sum(PD2.*[0.; diff(PF2)])
####################################################################
## Another ROC curve
# Julia code to generate the ROC curve.
using Distributions, Plots
σ = 2.
μ = 3.
PF1 = zeros(1000); PD1 = zeros(1000)
PF2 = zeros(1000); PD2 = zeros(1000)
αset = range(0.,0.5,length=1000)
d = Normal()
for i=1:1000
α = αset[i]
PF1[i] = 2. *α
PD1[i] = 1. - (cdf(d, quantile(d,1. - α) - μ/σ) -
cdf(d, -quantile(d, 1. -α) - μ/σ))
end
αset = range(0.,1.,length=1000)
for i=1:1000
α = αset[i]
PF2[i] = α
PD2[i] = 1. - cdf(d, quantile(d, 1-α)-μ/σ)
end
args = (linewidth=3, xlabel = "p_F", ylabel="p_D")
plot(PF1, PD1; args..., label = "Proposed decision")
plot!(PF2, PD2; args..., label = "Neyman-Pearson", legend=:bottomright)
####################################################################
## ROC on real data
using DelimitedFiles, Plots
download("https://probability4datascience.com/data/ch9_ROC_example_data.txt", "./ch9_ROC_example_data.txt")
scores = readdlm("ch9_ROC_example_data.txt")
scores = scores[:] # a vector
tau = range(0.,1.,length=1000)
labels = [ones(50); zeros(50)]
PF = zeros(1000)
PD = zeros(1000)
for i=1:1000
idx = (scores .<= tau[i])
predict = zeros(100)
predict[idx] .= 1.
true_positive = 0; true_negative = 0;
false_positive = 0; false_negative = 0;
for j=1:100
if (predict[j]==1) && (labels[j]==1)
true_positive += 1 end
if (predict[j]==1) && (labels[j]==0)
false_positive += 1 end
if (predict[j]==0) && (labels[j]==1)
false_negative += 1 end
if (predict[j]==0) && (labels[j]==0)
true_negative += 1 end
end
PF[i] = false_positive/50
PD[i] = true_positive/50
end
plot(PF, PD, linewidth=3, legend=false)