churn_sim/churn_2_crude-rate.R at main · jerudolph13/churn_sim · GitHub

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

packages <- c("tidyverse", "tableone", "survival", "broom")
for (package in packages){
  suppressPackageStartupMessages(library(package, character.only=T, quietly=T))
}

nsim <- 1000

for (model in c("dag1", "dag2.1", "dag2.2", "dag2.3", "dag2.4",
                "dag3", "dag3.1", "dag3.2", "dag3.3", "dag4", "dag4.1")) {
for (outcome in c("transient", "permanent", "repeated")) {


# Read in results ---------------------------------------------------------

dat <- read_csv(paste0("../data/", model, "_", outcome, ".csv"))


# Look at data ------------------------------------------------------------

#CreateTableOne(data=dat, vars=c("B1", "B2", "Z", "Y"), strata="M")


# Estimate rate -----------------------------------------------------------

rep.res <- function(r) {

  dat.r <- filter(dat, sim_rep==r)

  # True natural course
  true.dat <- dat.r %>%
    group_by(id) %>%
    mutate(cumY = cumsum(cumsum(Y)),
           last_t = lag(t)) %>%
    filter(cumY<=1) %>%
    filter(t>0)

  true.haz <- true.dat %>%
    group_by(t) %>%
    summarize(d = sum(Y),
              n = n()) %>%
    mutate(haz_truth = d/n)%>%
    select(t, haz_truth)
  true.rate <- sum(true.dat$Y)/sum(true.dat$t - true.dat$last_t)

  # Censor at missed visit
  obs.dat <- dat.r %>%
    filter(M==0)

  censor.dat <- obs.dat %>%
    group_by(id) %>%
    mutate(last_t = lag(t, default=0),
           diff_t = t - last_t,
           gap = diff_t>1,
           cum_gap = cumsum(gap)) %>%
    filter(cum_gap<1)

  censor.dat <- censor.dat %>%
    mutate(cumY = cumsum(cumsum(Y))) %>%
    filter(cumY<=1) %>%
    filter(t>0)

  censor.haz <- censor.dat %>%
    group_by(t) %>%
    summarize(d = sum(Y),
              n = n()) %>%
    mutate(haz_censor = d/n)%>%
    select(t, haz_censor)
  censor.rate <- sum(censor.dat$Y)/sum(censor.dat$t - censor.dat$last_t)

  # Allow participants to return after missing visit (use all time)
  gap.dat <- obs.dat %>%
    group_by(id) %>%
    mutate(cumY = cumsum(cumsum(Y)),
           last_t = lag(t)) %>%
    filter(cumY<=1) %>%
    filter(t>0)

  gap.haz1 <- gap.dat %>%
    group_by(t) %>%
    summarize(d = sum(Y),
              n = n()) %>%
    mutate(haz_gap1 = d/n)%>%
    select(t, haz_gap1)
  gap.rate1 <- sum(gap.dat$Y)/sum(gap.dat$t - gap.dat$last_t)

  # Allow participants to return after missing visit (exclude missed time)
  gap.dat <- obs.dat %>%
    group_by(id) %>%
    mutate(cumY = cumsum(cumsum(Y)),
           last_t = t-1) %>%
    filter(cumY<=1) %>%
    filter(t>0)

  gap.haz2 <- gap.dat %>%
    group_by(t) %>%
    summarize(d = sum(Y),
              n = n()) %>%
    mutate(haz_gap2 = d/n) %>%
    select(t, haz_gap2)
  gap.rate2 <- sum(gap.dat$Y)/sum(gap.dat$t - gap.dat$last_t)

  # Combine results
  res.haz <- data.frame(rep = r,
                    t = seq(1, 10, 1)) %>%
    left_join(true.haz, by="t") %>%
    left_join(censor.haz, by="t") %>%
    left_join(gap.haz1, by="t") %>%
    left_join(gap.haz2, by="t")

  res.rate <- bind_cols(rate_truth = true.rate,
                        rate_censor = censor.rate,
                        rate_gap1 = gap.rate1,
                        rate_gap2 = gap.rate2)

  res <- bind_cols(res.haz, res.rate)

  return(res)

}

all.res <- lapply(1:nsim, function(x){rep.res(x)})
all.res <- bind_rows(all.res)

write.csv(all.res, paste0("../results/", model, "_", outcome, "_rate_all.csv"))

summ.res <- all.res %>%
  group_by(t) %>%
  summarize(across(!rep, list(avg = ~mean(.x, na.rm=T), sd = ~sd(.x, na.rm=T)))) %>%
  mutate(bias_haz_truth = haz_truth_avg - haz_truth_avg,
         bias_haz_censor = haz_censor_avg - haz_truth_avg,
         bias_haz_gap_allt = haz_gap1_avg - haz_truth_avg,
         bias_haz_gap_1t = haz_gap2_avg - haz_truth_avg,

         bias_rate_truth = rate_truth_avg - rate_truth_avg,
         bias_rate_censor = rate_censor_avg - rate_truth_avg,
         bias_rate_gap_allt = rate_gap1_avg - rate_truth_avg,
         bias_rate_gap_1t = rate_gap2_avg - rate_truth_avg) %>%
  ungroup()

write.csv(summ.res, paste0("../results/", model, "_", outcome, "_rate_summ.csv"))


}}