-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtakeup_rct_target_villages.R
222 lines (195 loc) · 11.2 KB
/
takeup_rct_target_villages.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
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
# Identify valid target villages and clusters -----------------------------
rct.cluster.selection <- read_rds("rct_cluster_selection_2.0.rds")
rct.targetable.schools <- read_rds("rct_targetable_schools_2.0.rds")
clusters.to.drop <- c("99", "691", "892", "1293", "239") %>% # Problems with clusters at cluster survey stage; dropped from study
c("201", "853", "1402") # These are are study cluster that are no longer valid based on distance to other PoT
rct.cluster.selection %<>% magrittr::extract(!.$cluster.id %in% clusters.to.drop, )
rct.cluster.selection@data %<>% left_join(rct.schools.data %>%
spTransform(wgs.84) %>%
as.data.frame %>%
select(cluster.id, lon, lat)) %>%
rename(pot.lon = lon, pot.lat = lat)
rct.cluster.selection %<>%
`@`(data) %>%
left_join(cluster.survey.data %>%
distinct(cluster.id, found, num.valid.villages, gpslatitude, gpslongitude, gps2latitude, gps2longitude) %>%
filter(!duplicated(cluster.id)),
by = "cluster.id") %>%
mutate(alt.pot.lat = gps2latitude,
alt.pot.lon = gps2longitude,
pot.lat = ifelse(!is.na(gpslatitude), #& found %in% c("Found", "Name Changed"),
gpslatitude,
ifelse(!is.na(gps2latitude), #& found %in% c("Closed", "Not Found"),
alt.pot.lat,
pot.lat)),
pot.lon = ifelse(!is.na(gpslongitude), #& found %in% c("Found", "Name Changed"),
gpslongitude,
ifelse(!is.na(gps2longitude), #& found %in% c("Closed", "Not Found"),
alt.pot.lon,
pot.lon))) %>%
as.data.frame %>%
convert.to.sp(~ pot.lon + pot.lat, wgs.84) %>%
spTransform(kenya.proj4)
cluster.survey.data %<>%
filter(!is.na(target.lat), !is.na(target.lon)) %>%
arrange(cluster.id, target.village.id) %>%
left_join(filter(., !is.na(target.lon), !is.na(target.lat)) %>%
ddply(.(cluster.id), function(cluster.villages) {
convert.to.sp(cluster.villages, ~ target.lon + target.lat, wgs.84) %>%
spTransform(kenya.proj4) %>%
gDistance(rct.cluster.selection, byid = TRUE) %>%
set_colnames(cluster.villages$target.village.id) %>%
magrittr::extract(, !duplicated(colnames(.)), drop = FALSE) %>%
tibble::as_tibble() %>%
mutate(dist.from.cluster.id = rct.cluster.selection$cluster.id,
cluster.id = cluster.villages$cluster.id[1],
cluster.group = rct.cluster.selection$cluster.group) %>%
tidyr::gather(target.village.id, dist, -c(dist.from.cluster.id, cluster.id, cluster.group)) %>%
left_join(.,
select(., dist.from.cluster.id, target.village.id, dist),
by = c("cluster.id" = "dist.from.cluster.id", "target.village.id"),
suffix = c(".to.other.pot", ".to.pot")) %>%
filter(dist.from.cluster.id != cluster.id) %>%
select(-dist.from.cluster.id) %>%
group_by(target.village.id, cluster.group) %>%
filter(row_number(dist.to.other.pot) == 1) %>%
ungroup %>%
spread(cluster.group, dist.to.other.pot) %>%
mutate(valid.target.village = bracelet.airtime >= 3860 & control.ink >= 3000,
target.village.id = as.integer(target.village.id))
}),
by = c("cluster.id", "target.village.id")) %>%
left_join(filter(rct.targetable.schools, selected.targeted) %>%
select(pot.cluster.id, village.dist.cat, cluster.dist.cat),
by = c("cluster.id" = "pot.cluster.id")) %>%
group_by(cluster.id) %>% {
if (first(.$village.dist.cat) == "far") arrange(., desc(dist.to.pot)) else arrange(., dist.to.pot)
} %>%
filter(!duplicated(target.village_name)) %>%
ungroup %>%
mutate(target.actual.village.dist.cat = ifelse(dist.to.pot <= (2500/2), "close", "far"),
target.valid.dist.to.pot = target.actual.village.dist.cat == village.dist.cat,
target.in.range = dist.to.pot <= 2500,
dist.switchable = !target.valid.dist.to.pot & valid.target.village,
valid.target.village = valid.target.village & target.in.range,
target.village.pop.size = ifelse(target.pop_households <= median(target.pop_households, na.rm = TRUE), "small", "big")) %>%
filter(!is.na(target.village.pop.size)) %>%
group_by(cluster.id) %>%
mutate(cluster.pop.size.strata = unique(target.village.pop.size) %>% { ifelse(length(.) == 2, "mixed", .) },
actual.cluster.dist.cat = unique(target.actual.village.dist.cat) %>% { ifelse(length(.) == 2, "mixed", .) }) %>%
ungroup
rct.cluster.selection@data %<>% left_join(cluster.survey.data %>%
mutate(backup.cluster = FALSE) %>% # cluster.id %in% rct.backup.clusters$cluster.id) %>%
group_by(cluster.id, backup.cluster, cluster.pop.size.strata, actual.cluster.dist.cat) %>%
summarize(any.buffer.valid.village = any(valid.target.village),
dist.switchable = any(dist.switchable)) %>%
ungroup,
by = "cluster.id") %>%
left_join(rct.schools.data@data %>%
filter(selected.targeted) %>%
select(pot.cluster.id, village.dist.cat, cluster.dist.cat),
by = c("cluster.id" = "pot.cluster.id"))
rct.cluster.selection@data %<>%
mutate(village.dist.cat = ifelse(dist.switchable & !backup.cluster, ifelse(village.dist.cat == "close", "far", "close"), village.dist.cat),
any.buffer.valid.village = (dist.switchable & !backup.cluster) | any.buffer.valid.village)
# Display RCT clusters ----------------------------------------------------
rct.cluster.selection@data %>%
filter(!any.buffer.valid.village) %>%
select(cluster.id, cluster.group) %>%
left_join(cluster.survey.data) %>%
select(cluster.id, matches("valid"), bracelet.airtime, control.ink, dist.to.pot, village.dist.cat, cluster.group, dist.switchable)
# Village selection -------------------------------------------------------
rct.villages <- cluster.survey.data %>%
filter(valid.target.village, cluster.id %in% rct.cluster.selection$cluster.id) %>%
group_by(cluster.id) %>%
sample_n(1) %>%
ungroup
# write_rds(rct.villages, "rct_target_villages_2.0.rds")
rct.villages <- read_rds("rct_target_villages_2.0.rds")
unused.villages <- cluster.survey.data %>% anti_join(rct.villages, by = c("cluster.id", "target.village.id"))
# reassign.pop.size.strata <- . %>% {
# if (first(.$cluster.pop.size.strata) == "mixed" & first(.$actual.cluster.dist.cat != "mixed")) {
# stopifnot(first(.$missing.big.pop.size) <= nrow(.))
#
# new.strata.assignment <- sample(c(rep("big", first(.$missing.big.pop.size)), rep("small", nrow(.) - first(.$missing.big.pop.size))))
#
# mutate(., assigned.pop.size.strata = new.strata.assignment)
# } else return(.)
# }
#
# reassign.dist.cat <- . %>% {
# if (first(.$cluster.pop.size.strata) != "mixed" & first(.$actual.cluster.dist.cat == "mixed")) {
# stopifnot(first(.$missing.far.size) <= nrow(.))
#
# new.dist.assignment <- sample(c(rep("far", first(.$missing.far.size)), rep("close", nrow(.) - first(.$missing.far.size))))
#
# mutate(., assigned.cluster.dist.cat = new.dist.assignment)
# } else return(.)
# }
#
# reassign.mixed.mixed <- . %>% {
#
# }
#
# rct.cluster.selection@data %>%
# filter(any.buffer.valid.village) %>%
# # Assign strata to "mixed" clusters
# mutate(assigned.pop.size.strata = cluster.pop.size.strata,
# assigned.cluster.dist.cat = actual.cluster.dist.cat) %>%
# group_by(cluster.group) %>%
# mutate(cluster.group.size = n()) %>%
# group_by(actual.cluster.dist.cat, add = TRUE) %>%
# mutate(assigned.big.pop.size = n() %/% 2 + sample(seq(0, n() %% 2), 1),
# missing.big.pop.size = max(0, first(assigned.big.pop.size) - sum(cluster.pop.size.strata == "big"))) %>%
# group_by(cluster.pop.size.strata, add = TRUE) %>%
# do(reassign.pop.size.strata(.)) %>%
# group_by(cluster.group, cluster.pop.size.strata) %>%
# mutate(assigned.far.size = n() %/% 2 + sample(seq(0, n() %% 2), 1),
# missing.far.size = max(0, first(assigned.far.size) - sum(actual.cluster.dist.cat == "far"))) %>%
# group_by(actual.cluster.dist.cat, add = TRUE) %>%
# do(reassign.dist.cat(.)) %>%
# group_by(cluster.group) %>%
# ungroup %>%
# count(cluster.group, assigned.cluster.dist.cat, assigned.pop.size.strata)
# RCT target villages info table ------------------------------------------
write.rct.village.info <- function(.data, .file.name) {
.data %>%
select(cluster.id, target.village_name, target.lon, target.lat,
matches("boda"), matches("target.*elder"), matches("target.*chv")) %>%
left_join(rct.cluster.selection %>%
tidy %>%
select(cluster.id, school.name, county, constituency, ward, pot.lon, pot.lat),
by = "cluster.id") %>%
arrange(as.integer(cluster.id)) %>%
write_csv(.file.name, na = "")
}
rct.villages %>% write.rct.village.info("takeup_rct_villages.csv")
# Plot RCT clusters and villages ------------------------------------------
rct.cluster.selection %>%
magrittr::extract(.$any.buffer.valid.village, ) %>%
spTransform(wgs.84) %>%
tidy %>%
transmute(lon = pot.lon, lat = pot.lat, type = "PoT", cluster.id) %>%
bind_rows(rct.villages %>%
transmute(lon = target.lon, lat = target.lat, type = "Target Village", cluster.id)) %>% {
spatial.joint.data <- filter(., type == "PoT") %>%
convert.to.sp(~ lon + lat, wgs.84) %>% {
rbind(mutate(tidy(spTransform(buffer.clusters(., .width = 3000), wgs.84)), cluster.group = "control.ink"),
mutate(tidy(spTransform(buffer.clusters(., .width = 4000), wgs.84)), cluster.group = "bracelet.calendar"))
} %>%
mutate(id = paste(cluster.group, id, sep = "-"))
ggmap(get_map(make_bbox(lon, lat, data = .), maptype = "toner")) +
geom_point(aes(lon, lat, color = type), shape = 1, stroke = 1.5, alpha = 0.5, data = .) +
geom_line(aes(lon, lat, group = cluster.id), data = .) +
geom_polygon(aes(long, lat, group = id, fill = cluster.group), alpha = 0.25, data = spatial.joint.data) +
labs(x = "", y = "") +
scale_color_manual("", values = c("red", "darkgreen")) +
theme(legend.position = "bottom")
}
# Intercluster distance ---------------------------------------------------
rct.villages %>%
left_join(rct.cluster.selection@data %>% select(cluster.id, cluster.group)) %>%
group_by(cluster.group) %>%
do(dist.mat = convert.to.sp(., ~ target.lon + target.lat, wgs.84) %>%
spTransform(kenya.proj4) %>%
gDistance(rct.cluster.selection, byid = TRUE))