-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathOptimal_Lag.R
185 lines (171 loc) · 6.79 KB
/
Optimal_Lag.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
library(lavaan)
library(ggplot2)
library(semPlot)
library(lavaanPlot)
library(dplyr)
library(piecewiseSEM)
library(nlme)
library(lme4)
# Read DATA
Agg_Trips_1 <- read.csv('D:/COVID-19/PNAS_SECOND/All_XY_Features_To_R_County_Level_0731_toR.csv')
Agg_Trips_1 <- Agg_Trips_1[with(Agg_Trips_1, order(CTFIPS, Date)),]
Agg_Trips_1$Is_ReopenState <- TRUE
Agg_Trips_1$Date <- as.Date(Agg_Trips_1$Date)
Start_date <- as.Date('2020-03-10')
Max_Day <- as.numeric(difftime(max(Agg_Trips_1$Date), Start_date, units = 'days'))
time_window <- 7
# A function for all state
All_State_SEM <- function(Max_Day, Agg_Trips_1, time_window, model, xvar) {
# RUN LOOP
Start_date <- as.Date('2020-03-10')
All_corr_Reopen <- c()
All_Predict <- c()
for (jj in (1:(Max_Day - 7))) {
print(jj)
skip_to_next <- FALSE
Agg_Trips_tem <- Agg_Trips_1[(Agg_Trips_1$Date <= Start_date + time_window) &
(Agg_Trips_1$Date > Start_date) &
(Agg_Trips_1$New_cases > 0) &
(Agg_Trips_1$InFlow_Weight > 0),]
Agg_Trips_tem <- na.omit(Agg_Trips_tem)
rownames(Agg_Trips_tem) <- NULL
tryCatch({
fit <- sem(model, data = Agg_Trips_tem) # ,orthogonal = TRUE,std.lv = TRUE
#anova(fit, fit.partial)
para <- parameterEstimates(fit)
# Predict the cases
Beta <- para[(para$lhs == 'Log_New_cases') & (para$op == "~"), 'est']
xNames <- para[(para$lhs == 'Log_New_cases') & (para$op == "~"), 'rhs']
Agg_Trips_tem$pred.cases <- c(as.matrix(Agg_Trips_tem[xNames]) %*% Beta)
para$Date <- Start_date
All_corr_Reopen[[jj]] <- para
All_Predict[[jj]] <- Agg_Trips_tem
},
error = function(e) { skip_to_next <<- TRUE })
Start_date <- Start_date + 1
if (skip_to_next) { next }
}
All_corr_Reopen <- do.call(rbind.data.frame, All_corr_Reopen)
All_predict <- do.call(rbind.data.frame, All_Predict)
# PLOT COEFFI
All_corr_1_Reopen <- All_corr_Reopen[(All_corr_Reopen$lhs == 'Log_New_cases') &
(All_corr_Reopen$rhs == xvar) &
(All_corr_Reopen$Date > as.Date('2020-03-10')) &
(All_corr_Reopen$pvalue < 0.1),]
All_corr_1_Reopen <- na.omit(All_corr_1_Reopen)
rownames(All_corr_1_Reopen) <- NULL
# Print RESIDUAL
# Let <0 to 0
All_predict$pred.cases[All_predict$pred.cases < 0] <- 0
A_resi <- mean(abs(All_predict$pred.cases - All_predict$Log_New_cases))
list_result <- list(All_corr_Reopen, All_predict, All_corr_1_Reopen, A_resi)
return(list_result)
}
# LOOP IN LAG
# FINISH A LAG FIRST
All_resid <- c()
All_resid_result <- c()
for (jj in (0:30)) {
lag_data <- Agg_Trips_1 %>%
group_by(CTFIPS) %>%
mutate(lag.value = dplyr::lag(Log_Risked_WInput, n = jj, default = NA))
#lag_data <-lag_data[!rowSums(is.na(lag_data['lag.value15'])), ]
lag_data <- lag_data %>%
group_by(CTFIPS) %>%
filter((row_number() > 30))
All_corr_1 <- All_State_SEM(Max_Day, lag_data, time_window,
model = 'Log_New_cases ~ lag.value + Is_Weekend + Population_density + Pct_Age_0_24 + Pct_Age_25_40 + Pct_Age_40_65',
xvar = 'lag.value')
All_resid[[jj + 1]] <- All_corr_1[[4]]
All_corr_1[[2]]$LAG_NUM <- jj + 1
All_resid_result[[jj + 1]] <- select(All_corr_1[[2]], pred.cases, Log_New_cases, LAG_NUM)
}
All_resid1 <- do.call(rbind.data.frame, All_resid)
colnames(All_resid1) <- 'resi'
plot(All_resid1$resi)
All_resid2 <- do.call(rbind.data.frame, All_resid_result)
All_resid2$resid <- abs(All_resid2$pred.cases - All_resid2$Log_New_cases)
# Plot bar
boxplot(resid ~ LAG_NUM, data = All_resid2, outline = FALSE, range = 1)
Resid_summ <- All_resid2 %>%
group_by(LAG_NUM) %>%
summarise_each(funs(mean, median, sd))
ggplot(Resid_summ, aes(x = LAG_NUM, y = resid_mean)) +
geom_errorbar(aes(ymin = resid_mean - resid_sd * 0.05, ymax = resid_mean + resid_sd * 0.05), width = 0.5) +
geom_line() +
geom_point() +
labs(x = "Lag", y = "MAE") +
theme_bw()
ggplot(Resid_summ, aes(x = LAG_NUM, y = resid_mean)) +
geom_line() +
geom_point() +
theme_bw()
# TRY PANLE MODEL
# A function for all state
All_State_SEM_Panel_Lag <- function(Max_Day, Agg_Trips_1, time_window) {
# RUN LOOP
All_resid <- c()
Start_date <- as.Date('2020-03-10')
for (jj in (1:(Max_Day - 7))) {
print(jj)
skip_to_next <- FALSE
Agg_Trips_tem <- Agg_Trips_1[(Agg_Trips_1$Date <= Start_date + time_window) &
(Agg_Trips_1$Date > Start_date) &
(Agg_Trips_1$New_cases > 0) &
(Agg_Trips_1$InFlow_Weight > 0),]
Agg_Trips_tem <- na.omit(Agg_Trips_tem)
rownames(Agg_Trips_tem) <- NULL
tryCatch({
model.list <- list(
lme(Log_New_cases ~ lag.value +
Lag1_Log_New_cases +
Is_Weekend +
Population_density +
Pct_Age_0_24 +
Pct_Age_25_65 +
Med_House_Income, random = ~1 | CTFIPS, na.action = na.omit, data = Agg_Trips_tem))
fit <- as.psem(model.list) # ,orthogonal = TRUE,std.lv = TRUE
#new.summary <- summary(fit, .progressBar = F)
resi <- data.frame(residuals(fit))
#anova(fit, fit.partial)
All_resid[[jj]] <- resi
},
error = function(e) { skip_to_next <<- TRUE })
Start_date <- Start_date + 1
if (skip_to_next) { next }
}
All_resid <- do.call(rbind.data.frame, All_resid)
return(All_resid)
}
# LOOP IN LAG
# FINISH A LAG FIRST
All_resid <- c()
for (jj in (0:30)) {
lag_data <- Agg_Trips_1 %>%
group_by(CTFIPS) %>%
mutate(lag.value = dplyr::lag(Log_Risked_WInput, n = jj, default = NA))
#lag_data <-lag_data[!rowSums(is.na(lag_data['lag.value15'])), ]
lag_data <- lag_data %>%
group_by(CTFIPS) %>%
filter((row_number() > 30))
All_corr_1 <- All_State_SEM_Panel_Lag(Max_Day, lag_data, time_window)
All_corr_1$LAG_NUM <- jj + 1
All_resid[[jj + 1]] <- All_corr_1
}
All_resid1 <- do.call(rbind.data.frame, All_resid)
All_resid1$Log_New_cases_residuals <- abs(All_resid1$Log_New_cases_residuals)
# Plot bar
boxplot(Log_New_cases_residuals ~ LAG_NUM, data = All_resid1, outline = FALSE, range = 0.5)
Resid_summ <- All_resid1 %>%
group_by(LAG_NUM) %>%
summarise_each(funs(mean, median, sd))
ggplot(Resid_summ, aes(x = LAG_NUM, y = median)) +
geom_line() +
geom_point() +
theme_bw()
ggplot(Resid_summ, aes(x = LAG_NUM, y = median)) +
geom_errorbar(aes(ymin = median - sd, ymax = median + sd), width = 0) +
geom_line() +
geom_point() +
labs(x = "Date", y = "Coeff") +
theme_bw()