-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy path05_state_raking.R
249 lines (191 loc) · 10.4 KB
/
05_state_raking.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
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
rm(list=ls())
library(data.table)
library(ggplot2)
library(dplyr)
library(tidyr)
library(rgdal)
library(broom)
library(rgeos)
library(weights)
library(spdep)
library(proj4)
library(maptools)
library(urbnmapr)
library(covidcast)
library(viridis)
library(boot)
library(lme4)
#Filepaths
in.path = "~/data_counties/Data/Processed/"
out.path = "~/counties-project/Results/"
acs = readRDS(paste0(out.path,"080523_ACS_predictions.RDS"))
acs_sample_sizes = readRDS(paste0(in.path,"ACS_n_size.RDS"))
yrbs = readRDS(paste0(in.path,"YRBS_final.RDS"))
yrbs$age = factor(yrbs$age, levels = c(15,16,17,18,19)) ##Dunno why this isn't saving in data processing
acs_sample_sizes$age = factor(acs_sample_sizes$age, levels = c(15,16,17,18,19))
acs_state = acs[,.(goschol,sitecode=state, state_code, county_prediction, pop2019, county_fips, age, sex,weight)]
acs_state = merge(acs_state,acs_sample_sizes)
######################
### Main Estimates ###
######################
#Construct raking factor on in school population only
acs_state_school = acs_state[goschol == 1]
yrbs = yrbs[,.(sitecode, weight, hadsex, age, sex)]
##Strata, state average
county_wts = acs_state_school[,.(weight_c = sum(weight)), by = c("county_fips","age","sex")]
acs_state_school = merge(acs_state_school,county_wts, by = c("county_fips","age","sex"))
yrbs_strata = yrbs[,.(yrbs_state = weighted.mean(hadsex, weight), n_yrbs = .N), by = c("age","sex","sitecode")]
acs_strata = acs_state_school[,.(acs_state = weighted.mean(county_prediction,weight_c)), by = c("age","sex","sitecode","state_code")]
acs_strata = merge(acs_strata, acs_sample_sizes)
##Add all other states
predicted = readRDS(paste0(out.path,"/yrbs_pred_model/090623_prediction_results.RDS"))[,type := "predict"]
extrap = readRDS(paste0(out.path,"/extrapolation_model/090623_prediction_results.RDS"))[,type := "extrap"]
impute = readRDS(paste0(out.path,"/yrbs_impute_model/090623_prediction_results.RDS"))[,type := "impute"]
setnames(predicted,"state","sitecode")
setnames(extrap,"state","sitecode")
setnames(impute,"state","sitecode")
yrbs_strata = rbind(yrbs_strata[,type := "direct"], predicted, extrap, impute, fill=T)
#Create 'raking factor', i.e. ratio of state level estimates, only on 'in school' population
combined = merge(acs_strata, yrbs_strata, by = c("age","sex","sitecode"), all = TRUE)
combined[,scale_factor := acs_state/yrbs_state]
combined[,logit_scale_factor := logit(acs_state)/logit(yrbs_state)]
saveRDS(combined,paste0(out.path,"073123_ACS_YRBS_combined.RDS"))
#Adjust the county values according to scaling factor
setnames(acs,"state","sitecode")
acs = merge(acs, combined[,.(age,sex,sitecode,scale_factor,yrbs_state)], by = c("age","sex","sitecode"))
acs[,raked := county_prediction/scale_factor]
#County predictions
county_wts_all = acs_state[,.(weight_c = sum(weight)), by = c("county_fips","age","sex")]
county_dat = copy(acs)
county_dat = merge(county_dat,county_wts_all, by = c("county_fips","age","sex"),all.x=TRUE)
county_dat = county_dat[,.(county_raked = weighted.mean(raked,weight_c), ##Note that weighting here does not make any difference
county_unraked = weighted.mean(county_prediction,weight_c)),
by=c("sitecode", "county_fips","cntyname", "age","sex")]
saveRDS(county_dat,paste0(out.path,"073123_ACS_raked.RDS"))
#saveRDS(county_dat,paste0(out.path,"090822_ACS_raked.RDS")) #No weights for means and updated n_acs
county_ranges = county_dat %>% group_by(age,sex) %>% summarise(county_min = min(county_raked,na.rm=T),
county_max = max(county_raked,na.rm=T))
state_ranges = yrbs_strata %>% group_by(age,sex) %>% summarise(state_min = min(yrbs_state,na.rm=T),
state_max = max(yrbs_state,na.rm=T))
county_dat[,source := "Raked county estimate"]
yrbs_strata[,source := "State estimate"]
two = rbind(county_dat[,.(age,sex,value = county_raked,source)],
yrbs_strata[,.(age,sex,value = yrbs_state, source)])
ggplot(county_dat, aes(age,county_raked)) + geom_violin() + facet_wrap(~sex)
gg = ggplot(two, aes(age,value, fill=source)) + geom_violin() + facet_wrap(~sex) + theme_bw() +
labs(y = "Proportion who ever had sex", x = NULL) + scale_fill_viridis_d() +
theme(axis.text.x = element_text(size = 10, face="bold", colour = "black"),
axis.text.y = element_text(size = 12, colour = "black"),
axis.title.y = element_text(size = 16),
plot.title = element_text(size = 16),
strip.background = element_rect(fill="white"),
strip.text = element_text(size = 12, face="bold"),
plot.caption = element_text(size = 10),
legend.text = element_text(size = 10))
#####################
### Uncertainty #####
#####################
#library(rsample)
#load(paste0(in.path,"state_prediction.RData"))
#n=100
set.seed(12345)
#xx <- rsample::bootstraps(all_pred_new, times = n, strata="sitecode")
#100 resamples
# yrbs_resample <- lapply(xx$splits, function(x){
# x <- data.table(analysis(x))
# state_years_w_responses = unique(x[!is.na(q58), .(sitecode, year)])
# state_years_w_responses = state_years_w_responses[year == 2019]
# subset_dt = merge(x, state_years_w_responses, by = c("sitecode","year"))
# subset_dt = subset_dt[age %in% 4:7]
# subset_dt = subset_dt[!is.na(sex) & !is.na(age) & !is.na(q58)]
# subset_dt[,id := paste0(sitecode,age,sex)]
# subset_dt[, hadsex := factor(q58, levels = c(1,2), labels = c(1,0))] # Change Y to factor
# subset_dt[, hadsex := as.numeric(as.character(hadsex))] # Ensure 0 is the reference category
# subset_dt = subset_dt[,.(yrbs_state = weighted.mean(hadsex, weight)), by = c("age","sex","sitecode")]
# return(subset_dt)
# })
#saveRDS(yrbs_resample,paste0(in.path,"yrbs_resample.RData"))
yrbs_resample = readRDS(paste0(in.path,"yrbs_resample.RData"))
yrbs_cast = rbindlist(yrbs_resample)
yrbs_cast[,id := rep(1:100, each = 288)]
yrbs_cast = dcast(yrbs_cast, age + sex + sitecode ~ id, value.var = "yrbs_state")
colnames(yrbs_cast) <- c("age","sex","sitecode",paste0("pred_",1:100))
uncertainty = readRDS(paste0(out.path,"080523_ACS_uncertainty.RDS"))
uncertainty_state = readRDS(paste0(out.path,"080523_ACS_uncertainty_state.RDS"))
uncertainty_school = readRDS(paste0(out.path,"080523_ACS_uncertainty_in_school.RDS"))
#Add estimated states
predicted = readRDS(paste0(out.path,"/yrbs_pred_model/090623_uncertainty.RDS"))
extrap = readRDS(paste0(out.path,"/extrapolation_model/090623_uncertainty.RDS")) ###NEED TO LIMIT THIS TO EXTRAPOLATED STATES
impute = readRDS(paste0(out.path,"/yrbs_impute_model/090623_uncertainty.RDS"))
setnames(acs,"state","sitecode", skip_absent = T)
uncertainty <- merge(unique(acs[,.(sitecode,county_fips)]),uncertainty, by = c("county_fips"))
uncertainty_school <- merge(unique(acs[,.(sitecode,county_fips)]),uncertainty_school, by = c("county_fips"))
yrbs_cast[,age := as.numeric(as.character(factor(age, levels = c(4,5,6,7), labels = c(15,16,17,18))))]
yrbs_cast[,sex1 := ifelse(sex == 1,"female","male")]
yrbs_cast[,sex := sex1]
yrbs_cast[,sex1 := NULL]
split_states1 = split(yrbs_cast,yrbs_cast$sitecode)
predicted[, run := paste0("pred_",run)]
predicted1 = dcast(predicted[,.(sitecode = state, age , sex, yrbs_state, run )],
sitecode + age + sex ~ run, value.var = "yrbs_state" )
setnames(extrap,"state","sitecode")
setnames(impute,"state","sitecode")
split_states2 = split(predicted1,predicted1$sitecode)
split_states3 = split(impute,impute$sitecode)
split_states4 = split(extrap,extrap$sitecode[extrap$sitecode %in% c("DE","HI","IN","NJ","NM","TN","WY")])
split_states <- c(split_states1,split_states2,split_states3,split_states4)
bind_all = data.table()
for(a in 15:18){
out1 = lapply(lapply(split_states,get_rf_uncertainty,a,"female"),"[[",1)
out1 = rbindlist(out1)
bind_all = rbind(bind_all,out1)
}
for(a in 15:18){
out1 = lapply(lapply(split_states,get_rf_uncertainty,a,"male"),"[[",1)
out1 = rbindlist(out1)
bind_all = rbind(bind_all,out1)
}
all_dat = readRDS(paste0(out.path,"061223_ACS_raked.RDS"))
county_out[,age := factor(age)]
all_dat = merge(county_out, all_dat, by = c("county_fips","age","sex"))
ggplot(all_dat, aes(mean, county_raked)) + geom_point() + geom_abline(slope = 1, intercept = 0)
###################
# Compare Means #
###################
combined = readRDS(paste0(out.path,"061223_ACS_YRBS_combined.RDS"))
check = uncertainty_state[,5:ncol(uncertainty_state)]
melt_uncertainty = uncertainty_state %>% select(-replicate) %>% melt(id.vars = c("age","sex","state"))
est = melt_uncertainty %>% group_by(age,sex,state) %>% summarise(mean = mean(value)) %>% data.table()
est_original = combined[,.(age,sex,state = sitecode, acs_state)]
est[,age := factor(age)]
all = merge(est, est_original, by = c("age","sex","state"))
ggplot(all, aes(acs_state,mean)) + geom_point() +
geom_abline(slope = 1,intercept = 0) + facet_wrap(~age + sex, scales = "free")
est[age == 15 & sex == "female" & state == "AK", mean]
est_original[age == 15 & sex == "female" & state == "AK", acs_state]
wight_cols = c(paste0("REPWT",1:80))
tt = acs[age == 15 & sex == "female" & sitecode == "AK"]
tt = tt[,wight_cols,with=FALSE]
apply(tt,1,mean)
#Raking factor
yrbs_cast <- yrbs_cast[rep(seq_len(nrow(yrbs_cast)), 81),] #Expanding to facilitate uncertainty calc
yrbs_cast <- yrbs_cast[order(sitecode, sex, age)]
uncertainty_state1 <- copy(uncertainty_state)[state %in% unique(yrbs_cast$sitecode)]
uncertainty_state1 <- uncertainty_state1[order(state,sex,age)]
create_seq1 = seq(1,nrow(uncertainty_state)+81,by=81*4*2)
#Raked county predictions
uncertainty1 <- merge(unique(acs[,.(sitecode,county_fips)]),uncertainty, by = c("county_fips"))
uncertainty1 <- uncertainty1[sitecode %in% unique(yrbs_cast$sitecode)]
uncertainty1 <- uncertainty1[order(sitecode,sex,age)]
state = length(unique(yrbs_cast$sitecode))
raking_factors <- vector("list",state)
county_raked <- vector("list",state)
for(i in 1:1){
yrbs1 = yrbs_cast[create_seq1[i]:(create_seq1[i+1]-1),4:ncol(yrbs_cast)]
acs_state = uncertainty_state1[create_seq1[i]:(create_seq1[i+1]-1),5:ncol(uncertainty_state)]
acs_county = uncertainty1[create_seq1[i]:(create_seq1[i+1]-1),6:ncol(uncertainty1)]
out = as.matrix(acs_state,nrow = 81,ncol = 100)/yrbs1
raking_factors[[i]] <- cbind(uncertainty_state[create_seq1[i]:(create_seq1[i+1]-1),.(age,sex,state)],out)
out_final = as.matrix(acs_county,nrow = 81,ncol = 100)/out
county_raked[[i]] <- cbind(uncertainty1[create_seq1[i]:(create_seq1[i+1]-1),.(county_fips,age,sex,state)],out_final)
}