-
Notifications
You must be signed in to change notification settings - Fork 20
/
power.analysis.R
222 lines (183 loc) · 8.57 KB
/
power.analysis.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
#' A priori power calculator
#'
#' This function performs an \emph{a priori} power estimation of a meta-analysis
#' for different levels of assumed between-study heterogeneity.
#'
#' @usage power.analysis(d, OR, k, n1, n2, p = 0.05, heterogeneity = 'fixed')
#'
#' @param d The hypothesized, or plausible overall effect size of a treatment/intervention under study compared
#' to control, expressed as the standardized mean difference (\emph{SMD}). Effect sizes must be positive
#' numerics (i.e., expressed as positive effect sizes).
#' @param OR The hypothesized, or plausible overall effect size of a treatment/intervention under study compared
#' to control, expressed as the Odds Ratio (\emph{OR}). If both \code{d} and \code{OR} are specified, results
#' will only be computed for the value of \code{d}.
#' @param k The expected number of studies to be included in the meta-analysis.
#' @param n1 The expected, or plausible mean sample size of the treatment group in the studies to be included in the meta-analysis.
#' @param n2 The expected, or plausible mean sample size of the control group in the studies to be included in the meta-analysis.
#' @param p The alpha level to be used for the power computation. Default is \eqn{\alpha = 0.05}.
#' @param heterogeneity Which level of between-study heterogeneity to assume for the meta-analysis. Can be either
#' \code{"fixed"} for no heterogeneity/a fixed-effect model, \code{"low"} for low heterogeneity, \code{"moderate"}
#' for moderate-sized heterogeneity or \code{"high"} for high levels of heterogeneity. Default is \code{"fixed"}.
#'
#' @details While researchers conducting primary studies can plan the size of their sample
#' based on the effect size they want to find, the situation is a different in
#' meta-analysis, where one can only work with the published material.
#' However, researchers have some control over the number of studies they want to include in their
#' meta-analysis (e.g., through more leniently or strictly defined inclusion criteria).
#' Therefore, one can change the power to some extent by including more or less studies into
#' the meta-analysis. Conventionally, a power of \eqn{1-\beta = 0.8} is deemed sufficient to detect an existing effect.
#' There are four things one has to make assumptions about when assessing the power of a meta-analysis a priori.
#'
#' \itemize{
#' \item The number of included or includable studies
#' \item The overall size of the studies we want to include (are the studies in the field rather small or large?)
#' \item The effect size. This is particularly important, as assumptions have to be made
#' about how big an effect size has to be to still be clinically meaningful. One study calculated
#' that for interventions for depression, even effects as small as \emph{SMD}=0.24 may still
#' be meaningful for patients (Cuijpers et al. 2014). If the aim is to study negative effects of an
#' intervention (e.g., death or symptom deterioration), even very small effect sizes are extremely
#' important and should be detected.
#' \item The heterogeneity of our studies’ effect sizes, as this also affects the precision of the pooled estimate,
#' and thus its potential to find significant effects.
#' }
#'
#' The \code{power.analysis} function implements the formula by Borenstein et al. (2011) to calculate
#' the power estimate. Odds Ratios are converted to \code{d} internally before the power is estimated, and
#' are then reconverted.
#'
#' @references
#'
#' Harrer, M., Cuijpers, P., Furukawa, T.A, & Ebert, D. D. (2019).
#' \emph{Doing Meta-Analysis in R: A Hands-on Guide}. DOI: 10.5281/zenodo.2551803. \href{https://bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/power-analysis.html}{Chapter 13}
#'
#'Cuijpers, P., Turner, E.H., Koole, S. L., Van Dijke, A., & Smit, F. (2014).
#'What Is the Threshold for a Clinically Relevant Effect? The Case of Major Depressive Disorders.
#'\emph{Depression and Anxiety, 31}(5): 374–78.
#'
#'Borenstein, M., Hedges, L.V., Higgins, J.P.T. and Rothstein, H.R. (2011). Introduction to Meta-Analysis. John Wiley & Sons.
#'
#' @author Mathias Harrer & David Daniel Ebert
#'
#' @return Returns a \code{list} with two elements:
#'
#' \itemize{
#' \item \code{Plot}: A plot showing the effect size (x), power (y), estimated power (red point) and
#' estimated power for changing effect sizes (blue line). A dashed line at 80\% power is also provided as a visual threshold for sufficient power.
#' \item \code{Power}: The \strong{estimated power} of the meta-analysis, expressed as a value between 0 and 1 (i.e., 0\%-100\%).
#' }
#'
#' @export power.analysis
#'
#' @import ggplot2
#'
#' @seealso \code{\link{power.analysis.subgroup}}
#'
#' @examples
#'
#' # Example 1: Using SMD and fixed-effect model (no heterogeneity)
#' power.analysis(d=0.124, k=10, n1=50, n2=50, heterogeneity = 'fixed')
#'
#' # Example 2: Using OR and assuming moderate heterogeneity
#' pa = power.analysis(OR=0.77, k=12, n1=50, n2=50, heterogeneity = 'high')
#' summary(pa)
#'
#' # Only show plot
#' plot(pa)
power.analysis = function(d, OR, k, n1, n2, p = 0.05, heterogeneity = "fixed") {
odds = FALSE
if (missing(OR) & missing(d)) {
stop("Either 'd' or 'OR' must be provided.")
}
if (!(heterogeneity %in% c("fixed", "low", "moderate", "high"))) {
stop("'heterogeneity' must be either 'fixed', 'low', 'moderate', 'high'.")
}
# Cohen's d not provided: calculate from OR
if (missing(d)) {
odds = TRUE
d = log(OR) * (sqrt(3)/pi)
token1 = "log"
} else {
token1 = "no.log"
}
es = d
if (heterogeneity == "fixed") {
het.factor = 1
v.d = ((n1 + n2)/(n1 * n2)) + ((d * d)/(2 * (n1 + n2)))
v.m = v.d/k
lambda = (d/sqrt(v.m))
plevel = 1 - (p/2)
zval = qnorm(p = plevel, 0, 1)
power = 1 - (pnorm(zval - lambda)) + (pnorm(-zval - lambda))
token2 = "fixed"
}
if (heterogeneity == "low") {
het.factor = 1.33
v.d = ((n1 + n2)/(n1 * n2)) + ((d * d)/(2 * (n1 + n2)))
v.m = v.d/k
v.m = 1.33 * v.m
lambda = (d/sqrt(v.m))
plevel = 1 - (p/2)
zval = qnorm(p = plevel, 0, 1)
power = 1 - (pnorm(zval - lambda)) + (pnorm(-zval - lambda))
token2 = "low"
}
if (heterogeneity == "moderate") {
het.factor = 1.67
v.d = ((n1 + n2)/(n1 * n2)) + ((d * d)/(2 * (n1 + n2)))
v.m = v.d/k
v.m = 1.67 * v.m
lambda = (d/sqrt(v.m))
plevel = 1 - (p/2)
zval = qnorm(p = plevel, 0, 1)
power = 1 - (pnorm(zval - lambda)) + (pnorm(-zval - lambda))
token2 = "moderate"
}
if (heterogeneity == "high") {
het.factor = 2
v.d = ((n1 + n2)/(n1 * n2)) + ((d * d)/(2 * (n1 + n2)))
v.m = v.d/k
v.m = 2 * v.m
lambda = (d/sqrt(v.m))
plevel = 1 - (p/2)
zval = qnorm(p = plevel, 0, 1)
power = 1 - (pnorm(zval - lambda)) + (pnorm(-zval - lambda))
token2 = "high"
}
# Loop for data for plot
dvec = (1:1000)/1000
if (d > 1) {
dvec = (1:(d * 1000))/1000
}
powvect = vector()
for (i in 1:length(dvec)) {
d = dvec[i]
v.d = ((n1 + n2)/(n1 * n2)) + ((d * d)/(2 * (n1 + n2)))
v.m = v.d/k
v.m = het.factor * v.m
lambda = (d/sqrt(v.m))
plevel = 1 - (p/2)
zval = qnorm(p = plevel, 0, 1)
powvect[i] = 1 - (pnorm(zval - lambda)) + (pnorm(-zval - lambda))
}
# Generate plot
if (odds == FALSE) {
plotdat = as.data.frame(cbind(dvec, powvect))
plot = ggplot(data = plotdat, aes(x = dvec, y = powvect)) + geom_line(color = "blue", size = 2) +
annotate("point", x = es, y = power, color = "red", size = 5) + theme_minimal() + geom_hline(yintercept = 0.8,
color = "black", linetype = "dashed") + ylab("Power") + xlab("Effect size (SMD)")
} else {
dvecs = exp(dvec * (pi/sqrt(3)))
dvec.inv = exp(-dvec * (pi/sqrt(3)))
dvec = as.vector(rbind(dvec.inv, dvecs))
powvect = as.vector(rbind(powvect, powvect))
plotdat = as.data.frame(cbind(dvec, powvect))
plot = ggplot(data = plotdat, aes(x = dvec, y = powvect)) + geom_line(color = "blue", size = 2) +
annotate("point", x = exp(es * (pi/sqrt(3))), y = power, color = "red", size = 5) + theme_minimal() +
geom_hline(yintercept = 0.8, color = "black", linetype = "dashed") + ylab("Power") + xlab("Effect size (OR)") +
scale_x_log10()
}
return.list = list("Plot" = plot, "Power" = power)
class(return.list) = c("power.analysis", token1, token2)
invisible(return.list)
return.list
}