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
|
epi.sscohortc <- function(N = NA, irexp1 = 0.25, irexp0 = 0.10, pexp = NA, n, power, r = 1, design = 1, sided.test = 2, finite.correction = FALSE, nfractional = FALSE, conf.level = 0.95){
alpha.new <- (1 - conf.level) / sided.test
z.alpha <- qnorm(1 - alpha.new, mean = 0, sd = 1)
if (!is.na(irexp1) & !is.na(n) & !is.na(power)){
stop("Error: at least one of exposed, n and power must be NA.")
}
# Sample size:
if(!is.na(irexp1) & !is.na(irexp0) & is.na(n) & !is.na(power)){
# Sample size estimate. From Woodward p 405:
z.beta <- qnorm(power, mean = 0, sd = 1)
# Risk ratio:
lambda <- irexp1 / irexp0
# Odds ratio:
psi <- (irexp1 / (1 - irexp1)) / (irexp0 / (1 - irexp0))
pi <- irexp0
pc <- (pi * ((r * lambda) + 1)) / (r + 1)
p1 <- (r + 1) / (r * (lambda - 1)^2 * pi^2)
p2 <- z.alpha * sqrt((r + 1) * pc * (1 - pc))
p3 <- z.beta * sqrt((lambda * pi * (1 - (lambda * pi))) + (r * pi * (1 - pi)))
n0 <- p1 * (p2 + p3)^2
# Account for the design effect:
n0 <- n0 * design
# Finite population correction:
n <- ifelse(is.na(N), n0, (n0 * N) / (n0 + (N - 1)))
if(nfractional == TRUE){
n.crude <- n
n.exp1 <- n / (r + 1) * r
n.exp0 <- n / (r + 1) * 1
n.total <- n.exp1 + n.exp0
}
if(nfractional == FALSE){
n.crude <- ceiling(n)
n.exp1 <- ceiling(n / (r + 1)) * r
n.exp0 <- ceiling(n / (r + 1)) * 1
n.total <- n.exp1 + n.exp0
}
rval <- list(n.total = n.total, n.exp1 = n.exp1, n.exp0 = n.exp0, power = power, irr = lambda, or = psi)
}
# Power:
else
if(!is.na(irexp1) & !is.na(irexp0) & !is.na(n) & is.na(power)){
# Study power. From Woodward p 409:
# Risk ratio:
lambda <- irexp1 / irexp0
# Odds ratio:
psi <- (irexp1 / (1 - irexp1)) / (irexp0 / (1 - irexp0))
pi <- irexp0
pc <- (pi * ((r * lambda) + 1)) / (r + 1)
# Account for the design effect:
n <- n / design
if(nfractional == TRUE){
n.exp1 <- (n / (r + 1)) * r
n.exp0 <- (n / (r + 1)) * 1
n.total <- n.exp1 + n.exp0
}
if(nfractional == FALSE){
n.exp1 <- ceiling((n / (r + 1)) * r)
n.exp0 <- ceiling((n / (r + 1)) * 1)
n.total <- n.exp1 + n.exp0
}
# Convert n (finite corrected sample size) to n0:
n0 <- ifelse(!is.na(N), (n * N - n) / (N - n), n)
t1 <- ifelse(lambda >= 1,
(pi * (lambda - 1) * sqrt(n0 * r)),
(pi * (1 - lambda) * sqrt(n0 * r)))
t2 <- z.alpha * (r + 1) * sqrt(pc * (1 - pc))
t3 <- (r + 1) * (lambda * pi * (1 - lambda * pi) + r * pi * (1 - pi))
z.beta <- (t1 - t2) / sqrt(t3)
power <- pnorm(z.beta, mean = 0, sd = 1)
rval <- list(n.total = n.total, n.exp1 = n.exp1, n.exp0 = n.exp0, power = power, irr = lambda, or = psi)
}
# Lambda:
else
if(is.na(irexp1) & !is.na(irexp0) & !is.na(n) & !is.na(power)){
# Risk ratio to be detected - requires a value for unexposed. From Woodward p 409:
z.beta <- qnorm(power, mean = 0, sd = 1)
pi <- irexp0
# Account for the design effect:
n <- n / design
n.exp1 <- (n / (r + 1)) * r
n.exp0 <- (n / (r + 1)) * 1
n.total <- n.exp1 + n.exp0
# Convert n (finite corrected sample size) to n0:
n0 <- ifelse(!is.na(N), (n * N - n) / (N - n), n)
Y <- r * n0 * pi^2
Z <- (r + 1) * pi * (z.alpha + z.beta)^2
a <- Y + (pi * Z)
b <- (2 * Y) + Z
c <- Y - (r * (1 - pi) * Z)
# Risk ratio:
lambda.pos <- (1 / (2 * a)) * (b + sqrt(b^2 - 4 * a * c))
lambda.neg <- (1 / (2 * a)) * (b - sqrt(b^2 - 4 * a * c))
rlambda.pos <- lambda.pos
rlambda.neg <- ifelse(lambda.neg < 0, 0, lambda.neg)
# From http://www.epigear.com/index_files/or2rr.html:
# s = prevalence of disease in the population
# p = prevalence of exposure in the population
# Prevalence of disease in the exposed, unexposed and population:
irexp1.pos <- lambda.pos * irexp0
irexp0.pos <- irexp0
s.pos <- (irexp1.pos + irexp0.pos) / 2
p.pos <- pexp
irexp1.neg <- lambda.neg * irexp0
irexp0.neg <- irexp0
s.neg <- (irexp1.neg + irexp0.neg) / 2
p.neg <- pexp
# Odds ratio:
psi.pos <- (lambda.pos * (1 - (s.pos / (p.pos * lambda.pos + 1 - p.pos)))) /
(1 - ((lambda.pos * s.pos) / (p.pos * lambda.pos + 1 - p.pos)))
psi.neg <- (lambda.neg * (1 - (s.neg / (p.neg * lambda.neg + 1 - p.neg)))) /
(1 - ((lambda.neg * s.neg) / (p.neg * lambda.neg + 1 - p.neg)))
rpsi.pos <- psi.pos
rpsi.neg <- ifelse(psi.neg < 0, 0, psi.neg)
rval <- list(n.total = n.total, n.exp1 = n.exp1, n.exp0 = n.exp0, power = power, irr = sort(c(rlambda.neg, rlambda.pos)), or = sort(c(rpsi.neg, rpsi.pos)))
}
rval
}
# epi.sscohortc(irexp1 = 0.25, irexp0 = 0.10, n = NA, power = 0.80, r = 1, design = 1, sided.test = 2, conf.level = 0.95)
# epi.sscohortc(irexp1 = 0.25, irexp0 = 0.10, n = 200, power = NA, r = 1, design = 1, sided.test = 2, conf.level = 0.95)
# epi.sscohortc(irexp1 = NA, irexp0 = 0.10, n = 200, power = 0.80, r = 1, design = 1, sided.test = 2, conf.level = 0.95)
|
