-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest-ridge.R
193 lines (161 loc) · 6.97 KB
/
test-ridge.R
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
rm(list = ls())
setwd("~/Dropbox/master/algo")
source("main-ridge.R")
library(tidyverse)
sitka10 <- read_table2("~/Dropbox/master/algo/data/sitka.txt") %>%
filter(id.num >= 1 & id.num <= 10) %>%
mutate(id.num = as.factor(id.num)) %>%
select(days, log.size, id.num)
growth10boys <- fda::growth[-3] %>%
map_dfc(function(x) as.tibble(x, rownames = "age")) %>%
gather(sub, height, -age, -age1) %>%
filter(grepl("^boy0[1-9]|^boy10", sub)) %>%
mutate(age = as.numeric(age), sub = as.factor(sub)) %>%
select(age, height, sub)
data <- growth10boys
deg <- 2
## 'x' invariant under scale and location transformation
## 'y' differ by a scale after scale transformation
x <- data[[1]] / max(data[[1]])
y <- data[[2]] / max(data[[2]])
## number of interior knots
K <- 8
## convert the group variable into a factor
if (is.factor(data[[3]])) {
grp <- droplevels(data[[3]])
} else {
grp <- factor(data[[3]], levels = unique(data[[3]]))
}
source("~/Dropbox/master/algo/subor.R")
## design matrices for the population curve (b-spline with transformation)
n_bsf <- K + deg + 1 # number of b-spline basis functions
D <- get_diff_mat(n_bsf, deg + 1) # difference matrix
des_ls_pop <- get_design_bs(x, K, deg)
G_pop <- cbind(1, poly(1:n_bsf, deg = deg, raw = FALSE, simple = TRUE))
H_pop <- crossprod(D, solve(tcrossprod(D)))
B_pop <- des_ls_pop$design %*% cbind(G_pop, H_pop)
K_mat <- cbind(matrix(0, K, deg + 1), diag(K))
source("main-ridge.R")
set.seed(1)
fm1 <- bayes_ridge(y, cbind(X_pop, Z_pop), K_mat, 0, 10000)
source("main-ridge.R")
set.seed(24)
system.time(fm2_1 <- bayes_ridge_sub(y, grp, B_pop, K_mat, deg + 1, 1000, 2000))
system.time(fm2_2 <- bayes_ridge_sub(y, grp, B_pop, K_mat, deg + 1, 1000, 5000))
system.time(fm2_3 <- bayes_ridge_sub(y, grp, B_pop, K_mat, deg + 1, 1000, 5000))
system.time(fm2_4 <- bayes_ridge_sub(y, grp, B_pop, K_mat, deg + 1, 1000, 5000))
fm2_2$basis <- list(type = 'bs_hier', knots = des_ls_pop$knots, degree = deg,
trans_mat = cbind(G_pop, H_pop))
fm2_2$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
source("subor.R")
plot_spline(fm2_2)
plot(fm2$samples$precision$pop)
acf(fm2$samples$precision$pop)
source("diagnostic.R")
ess_rfun(flatten_chains(fm2_1, fm2_2)[, , "sig2_delta2"])
source("main-ridge.R")
A_mat <- get_constmat_bs(NCOL(B_pop), "increasing") %*% cbind(G_pop, H_pop)
set.seed(2)
fm2c <- bayes_ridge_cons_sub(y, grp, B_pop, K_mat, deg + 1, A_mat, 0, 5000)
fm2c$basis <- list(type = 'bs_hier', knots = des_ls_pop$knots, degree = deg,
trans_mat = cbind(G_pop, H_pop))
fm2c$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
plot_spline(fm2c)
plot(fm2c$samples$population[1, ])
source("main-mc.R")
lower <- rep(0, NROW(A_mat))
fm2cmc <- mc_cons_sub(y, grp, B_pop, K_mat, A_mat, lower, fm2_1$samples$prec, 1000)
fm2cmc <- mc_cons_sub(y, grp, B_pop, K_mat, A_mat, lower, fm2_1$samples$prec, 1000)
fm2cmc$basis <- list(type = 'bs_hier', knots = des_ls_pop$knots, degree = deg,
trans_mat = cbind(G_pop, H_pop))
fm2cmc$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
plot_spline(fm2cmc)
source("main-mc.R")
set.seed(21)
system.time(fm2_1_re <- mc_sub(y, grp, cbind(X_pop, Z_pop), K_mat, fm2_1$samples$precision, 1000))
fm2_1_re$basis <- list(type = 'bs_hier', knots = des_ls_pop$knots, degree = deg,
trans_mat = cbind(G_pop, H_pop))
fm2_1_re$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
source("graphs.R")
plot_spline(fm2_1_re)
effectiveSize(mcmc(1 / fm2_1_re$samples$precision$eps))
effectiveSize(mcmc(1 / fm2_1_re$samples$precision$pop))
effectiveSize(mcmc(1 / fm2_1_re$samples$precision$sub2))
effectiveSize(mcmc(t(fm2_1_re$samples$population)))
plot(mcmc(t(fm2_1_re$samples$subject[, 2, ])))
## design matrices for the population curve (b-spline)
n_bsf <- K + deg + 1 # number of b-spline basis functions
des_ls_pop <- get_design_bs(x, K, deg)
B_pop <- des_ls_pop$design
K_mat <- get_diff_mat(n_bsf, deg + 1)
source("main-ridge.R")
set.seed(1)
fm3 <- bayes_ridge(y, B_pop, K_mat, 0, 5000)
plot_x <- seq(min(x), max(x), len = 100)
plot_y <- get_design_bs(plot_x, des_ls_pop$knots, deg)$design %*% fm3$means$population
ggplot() + geom_point(aes(x, y, col = grp)) + geom_line(aes(plot_x, plot_y))
source("main-ridge.R")
set.seed(2)
fm4 <- bayes_ridge_sub(y, grp, B_pop, K_mat, deg + 1, 0, 500)
fm4$basis <- list(type = 'bs', knots = des_ls_pop$knots, degree = deg)
fm4$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
source("graphs.R")
plot_spline(truncate_spline(fm4, 400))
source("main-ridge.R")
A_mat <- get_constmat_bs(NCOL(B_pop), "increasing")
set.seed(2)
fm4c <- bayes_ridge_cons_sub(y, grp, B_pop, K_mat, deg + 1, A_mat, 0, 10000)
fm4c$basis <- list(type = 'bs', knots = des_ls_pop$knots, degree = deg)
fm4c$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
plot_spline(fm4c)
plot(fm4c$samples$population[2, ], ylim = c(-2.15, -2.3))
plot(fm4c$samples$precision$pop)
plot(fm4c$samples$precision$sub1[1, 1, ])
plot(fm4c$samples$precision$sub2)
source("main-mc.R")
set.seed(21)
system.time(fm4_re <- mc_sub(y, grp, B_pop, K_mat, fm4$samples$precision, 500))
fm4_re$basis <- list(type = 'bs', knots = des_ls_pop$knots, degree = deg)
fm4_re$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
source("graphs.R")
plot_spline(fm4_re)
fm_stan <- readRDS("../stan/growth/bs-ridge.rds")
shinystan::launch_shinystan(fm_stan)
## design matrices for the population curve (tpf)
deg <- 2
des_ls_pop <- get_design_tpf(x, K, deg)
B_pop <- des_ls_pop$design
K_mat <- cbind(matrix(0, K, deg + 1), diag(K))
source("main-ridge.R")
set.seed(1)
fm5 <- bayes_ridge(y, B_pop, K_mat, 0, 10000)
plot_x <- seq(min(x), max(x), len = 100)
plot_y <- get_design_tpf(plot_x, des_ls_pop$knots, deg)$design %*% fm5$means$population
ggplot() + geom_point(aes(x, y, col = grp)) + geom_line(aes(plot_x, plot_y))
source("main-ridge.R")
set.seed(2)
system.time(fm6 <- bayes_ridge_sub(y, grp, B_pop, K_mat, deg + 1, 0, 5000))
fm6$basis <- list(type = 'tpf', knots = des_ls_pop$knots, degree = deg)
fm6$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
source("graphs.R")
plot_spline(fm6)
source("main-ridge.R")
A_mat <- get_constmat_tpf(des_ls_pop$knots, "increasing", deg)
set.seed(2)
fm6c <- bayes_ridge_cons_sub(y, grp, B_pop, K_mat, deg + 1, A_mat, 0, 5000)
plot(fm6c$samples$population[1, ])
plot(fm6c$samples$population[2, ])
plot(fm6c$samples$subjects[1, 1, ])
plot(1 / fm6c$samples$precision$pop, type = 'l')
plot(1 / fm6c$samples$precision$sub2)
fm6c$basis <- list(type = 'tpf', knots = des_ls_pop$knots, degree = deg)
fm6c$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
plot_spline(fm6c)
source("main-mc.R")
set.seed(21)
system.time(fm6_re <- mc_sub(y, grp, B_pop, K_mat, fm6$samples$precision, 1000))
fm6_re$basis <- list(type = 'tpf', knots = des_ls_pop$knots, degree = deg)
fm6_re$data <- data.frame(x = x, y = y, grp_sub = grp, grp_pop = NA)
source("graphs.R")
plot_spline(fm6_re)
plot(fm6_re$samples$subjects[1, 9, ])