# Originally based on https://github.com/richarddmorey/jpower
.runTest1Var <- function(jaspResults, options) {
stats <- .prepareStats(options)
## Compute results
results <- try(.computeTest1Var(jaspResults, options, stats))
if (inherits(results, "try-error")) {
.quitAnalysis(gettext("Unable to compute the power results. Try to enter less extreme values for the input parameters."))
}
.initPowerTabTest1Var(jaspResults, options, results, stats)
if (options$text) {
.initPowerESTabTest1Var(jaspResults, options, results, stats)
}
## Populate tables and plots
if (options$text) {
.populateIntro(jaspResults, options)
}
if (options$powerContour) {
.preparePowerContourTest1Var(jaspResults, options, results, stats)
if (options$text) {
.populateContourTextTest1Var(jaspResults, options, results, stats)
}
}
if (options$powerByEffectSize) {
.preparePowerCurveESTest1Var(jaspResults, options, results, stats)
if (options$text) {
.populatePowerCurveESTextTest1Var(jaspResults, options, results, stats)
}
}
if (options$powerBySampleSize) {
.preparePowerCurveNTest1Var(jaspResults, options, results, stats)
if (options$text) {
.populatePowerCurveNTextTest1Var(jaspResults, options, results, stats)
}
}
if (options$powerDemonstration) {
.preparePowerDistTest1Var(jaspResults, options, results, stats)
if (options$text) {
.populateDistTextTest1Var(jaspResults, options, results, stats)
}
}
if (options$saveDataset && options$savePath != "") {
.generateDatasetTest1Var(jaspResults, options, results, stats)
}
}
#### Compute results ----
.computeTest1Var <- function(jaspResults, options, stats) {
## Compute numbers for table
pow.n <- NULL
pow.es <- NULL
pow.pow <- NULL
if (options$calculation == "sampleSize") {
pow.n <- ceiling(.pwrVarTest(rho = stats$es, sig.level = stats$alpha, power = stats$pow, alternative = stats$alt)$n)
}
if (options$calculation == "effectSize") {
pow.es <- .pwrVarTest(n = stats$n, power = stats$pow, sig.level = stats$alpha, alternative = stats$alt)$rho
}
if (options$calculation == "power") {
pow.pow <- .pwrVarTest(n = stats$n, rho = stats$es, sig.level = stats$alpha, alternative = stats$alt)$power
}
return(list(n = pow.n, es = pow.es, power = pow.pow))
}
#### Init table ----
.initPowerTabTest1Var <- function(jaspResults, options, results, stats) {
table <- jaspResults[["powertab"]]
if (is.null(table)) {
# Create table if it doesn't exist yet
table <- createJaspTable(title = gettext("A Priori Power Analysis"))
table$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio"
))
table$position <- 2
jaspResults[["powertab"]] <- table
} else {
return()
}
calc <- options$calculation
if (calc == "sampleSize") {
order <- c(1, 2, 3, 4)
} else if (calc == "effectSize") {
order <- c(2, 1, 3, 4)
} else if (calc == "power") {
order <- c(3, 1, 2, 4)
} else {
order <- c(4, 1, 2, 3)
}
colNames <- c("sampleSize", "effectSize", "power", "alpha")
colLabels <- c(
"N",
gettext("Variance ratio (\u03C1)"),
gettext("Power"),
"\u03B1"
)
colType <- c("integer", "number", "number", "number")
for (i in seq_along(order)) {
table$addColumnInfo(colNames[order[i]],
title = colLabels[order[i]],
overtitle = if (i > 1) gettext("User Defined") else NULL,
type = colType[order[i]]
)
}
row <- list()
for (i in 2:4) {
row[[colNames[order[i]]]] <- options[[colNames[order[i]]]]
}
table$addRows(rowNames = 1, row)
.populatePowerTabTest1Var(jaspResults, options, results, stats)
}
.initPowerESTabTest1Var <- function(jaspResults, options, results, stats) {
table <- jaspResults[["powerEStab"]]
if (is.null(table)) {
# Create table if it doesn't exist yet
table <- createJaspTable(title = gettext("Power by variance ratio"))
table$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio",
"text"
))
table$position <- 4
jaspResults[["powerEStab"]] <- table
} else {
return()
}
table$addColumnInfo(
name = "es",
title = gettext("True variance ratio (\u03C1)"),
type = "string"
)
table$addColumnInfo(
name = "power",
title = gettext("Power to detect"),
type = "string"
)
table$addColumnInfo(
name = "desc",
title = gettext("Description"),
type = "string"
)
pow <- c("\u226450%", "50% \u2013 80%", "80% \u2013 95%", "\u226595%")
desc <- c(
gettext("Likely miss"),
gettext("Good chance of missing"),
gettext("Probably detect"),
gettext("Almost surely detect")
)
for (i in 1:4) {
row <- list("power" = pow[i], "desc" = desc[i])
table$addRows(rowNames = i, row)
}
.populatePowerESTabTest1Var(jaspResults, options, results, stats)
}
#### Populate texts ----
.populateContourTextTest1Var <- function(jaspResults, options, r, lst) {
html <- jaspResults[["contourText"]]
if (is.null(html)) {
html <- createJaspHtml()
html$dependOn(c("test", "text", "powerContour"))
html$position <- 6
jaspResults[["contourText"]] <- html
}
str <- gettextf(
"<p>The power contour plot shows how the sensitivity of the test changes with the hypothetical variance ratio and the sample sizes in the design. As we increase the sample sizes, smaller variance ratios become reliably detectable.<p>Conversely, if one is satisfied to reliably detect only larger variance ratios, smaller sample sizes are needed. The point shows the power of specified design and variance ratio."
)
html[["text"]] <- str
}
.populatePowerCurveESTextTest1Var <- function(jaspResults, options, r, lst) {
html <- jaspResults[["curveESText"]]
if (is.null(html)) {
html <- createJaspHtml()
html$dependOn(c("test", "text", "powerByEffectSize"))
html$position <- 8
jaspResults[["curveESText"]] <- html
}
## Get options from interface
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
power <- ifelse(calc == "power", r$power, lst$pow)
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
if (alt == "two.sided" && d < 1) {
d <- ifelse(calc == "effectSize",
d,
ifelse(calc == "sampleSize",
try(.pwrVarTest(n = n, power = power, sig.level = alpha, alternative = alt)$rho[2]),
lst$es
)
)
} else {
d <- ifelse(calc == "effectSize",
d,
ifelse(calc == "sampleSize",
try(.pwrVarTest(n = n, power = power, sig.level = alpha, alternative = alt)$rho[1]),
lst$es
)
)
}
if (inherits(d, "try-error")) {
return()
} else {
d <- round(as.numeric(d), 3)
}
n_text <- gettextf("sample sizes of %1$s", n)
if (d > 1) {
alt_text <- "<i>\u03C1\u003E</i>"
} else {
alt_text <- "<i>\u03C1\u003C</i>"
}
if (calc == "power") {
pwr_string <- gettextf("have power of at least %1$s", round(power, 3))
} else {
pwr_string <- gettextf("only be sufficiently sensitive (power > %1$s)", round(power, 3))
}
if (alt == "two.sided" && d < 1) {
d50 <- try(.pwrVarTest(n = n, sig.level = alpha, power = .5, alternative = alt)$rho[2])
if (inherits(d50, "try-error")) {
return()
}
interval <- gettextf("1 > %1$s > %2$s", "\u03C1", round(d50, 3))
} else {
d50 <- try(.pwrVarTest(n = n, sig.level = alpha, power = .5, alternative = alt)$rho[1])
if (inherits(d50, "try-error")) {
return()
}
if (alt == "less") {
interval <- gettextf("1 > %1$s > %2$s", "\u03C1", round(d50, 3))
} else {
interval <- gettextf("1 < %1$s < %2$s", "\u03C1", round(d50, 3))
}
}
str <- gettextf(
"<p>The power curve above shows how the sensitivity of the test and design is larger for larger variance ratios. If we obtained %1$s our test and design would %2$s to variance ratios of %3$s%4$s. <p>We would be more than likely to miss (power less than 50%%) variance ratios of %5$s.",
n_text, pwr_string, alt_text, d, interval
)
html[["text"]] <- str
}
.populatePowerCurveNTextTest1Var <- function(jaspResults, options, r, lst) {
html <- jaspResults[["curveNText"]]
if (is.null(html)) {
html <- createJaspHtml()
html$dependOn(c("test", "text", "powerBySampleSize"))
html$position <- 10
jaspResults[["curveNText"]] <- html
}
## Get options from interface
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
d <- round(d, 3)
power <- ifelse(calc == "power", r$power, lst$pow)
alt <- lst$alt
n_text <- gettextf("sample sizes of at least %1$s", n)
if (alt == "two.sided") {
alt_text <- "<i>\u03C1\u2260</i>1"
} else {
if (d > 1) {
alt_text <- "<i>\u03C1\u003E</i>1"
} else {
alt_text <- "<i>\u03C1\u003C</i>1"
}
}
str <- gettextf(
"<p>The power curve above shows how the sensitivity of the test and design is larger for larger sample sizes. In order for our test and design to have sufficient sensitivity (power > %1$s) to detect that %2$s when the variance ratio is %3$s or more extreme, we would need %4$s.",
round(power, 3), alt_text, d, n_text
)
html[["text"]] <- str
}
.populateDistTextTest1Var <- function(jaspResults, options, r, lst) {
html <- jaspResults[["distText"]]
if (is.null(html)) {
html <- createJaspHtml()
html$dependOn(c("test", "text", "powerDemonstration"))
html$position <- 12
jaspResults[["distText"]] <- html
}
## Get options from interface
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
d <- round(d, 2)
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
power <- ifelse(calc == "power",
r$power,
ifelse(calc == "sampleSize",
.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power,
lst$pow
)
)
n_text <- gettextf("a sample size of %1$s", n)
if (alt == "two.sided") {
tail_text <- gettext("two-sided")
null_text <- gettext("<i>\u03C1=</i>1,")
crit_text <- gettext("criteria")
hypo_text <- gettextf("%1$s%2$s", "\u03C1", "\u2260")
if (d > 1) {
alt_text <- "<i>\u03C1\u2265</i>"
}
if (d < 1) {
alt_text <- "<i>\u03C1\u2264</i>"
}
} else {
tail_text <- gettext("one-sided")
crit_text <- gettext("criterion")
if (d > 1) {
alt_text <- "<i>\u03C1\u2265</i>"
hypo_text <- gettextf("%1$s>", "\u03C1")
null_text <- "<i>\u03C1\u2264</i>1,"
}
if (d < 1) {
alt_text <- "<i>\u03C1\u2264</i>"
hypo_text <- gettextf("%1$s<", "\u03C1")
null_text <- "<i>\u03C1\u2265</i>1,"
}
}
str <- paste(
"<p>",
gettextf("The figure above shows two sampling distributions: the sampling distribution of the <i>estimated</i> variance ratio when <i>%1$s=</i>1, and when <i>%2$s=</i>%3$s.", "\u03C1", "\u03C1", d),
gettextf("Both assume %1$s.", n_text),
"</p><p>",
gettextf("The vertical dashed lines show the %1$s we would set for a %2$s test with <i>\u03B1=</i>%3$s.", crit_text, tail_text, alpha),
gettextf("When the observed variance ratio is far enough away from 1 to be more extreme than the %1$s we say we 'reject' the null hypothesis.", crit_text),
gettextf("If the null hypothesis were true and %1$s the evidence would lead us to wrongly reject the null hypothesis at most %2$s%% of the time.", null_text, 100 * alpha),
"</p><p>",
gettextf("On the other hand, if <i>%1$s</i>%2$s, the evidence would exceed the criterion — and hence we would correctly claim that <i>%3$s</i>1 — at least %4$s%% of the time.", alt_text, d, hypo_text, 100 * round(power, 3)),
gettextf("The design's power for detecting effects of %1$s%2$s is thus %3$s.", alt_text, d, round(power, 3)),
"</p>"
)
html[["text"]] <- str
}
#### Populate table ----
.populatePowerTabTest1Var <- function(jaspResults, options, r, lst) {
table <- jaspResults[["powertab"]]
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
power <- ifelse(calc == "power", r$power, lst$pow)
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
if (calc == "effectSize") {
row <- list()
row[[calc]] <- d
table$addColumns(row)
} else {
row <- list()
row[[calc]] <- r[[switch(calc,
"sampleSize" = "n",
calc
)]]
table$addColumns(row)
}
if (calc == "sampleSize") {
if (round(.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power, 3) == 1) {
table$addFootnote(gettextf("Due to the rounding of the sample size, the actual power can deviate from the target power. <b>Actual power: >0.999"))
} else {
table$addFootnote(gettextf(
"Due to the rounding of the sample size, the actual power can deviate from the target power. <b>Actual power: %1$s</b>",
round(.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power, 3)
))
}
}
}
.populatePowerESTabTest1Var <- function(jaspResults, options, r, lst) {
html <- jaspResults[["tabText"]]
if (is.null(html)) {
html <- createJaspHtml()
html$dependOn(c("test", "text"))
html$position <- 3
jaspResults[["tabText"]] <- html
}
## Get options from interface
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
power <- ifelse(calc == "power", r$power, lst$pow)
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
n_text <- gettextf("a sample size of ", n)
tail_text <- ifelse(alt == "two.sided",
gettext("two-sided"),
gettext("one-sided")
)
sign <- ifelse(d > 1, "\u2265", "\u2264")
if (calc == "sampleSize") {
str <- gettextf(
"We would need %1$s to reliably (with probability greater than %2$s) detect a variance ratio of <i>%3$s%4$s</i>%5$s, assuming a %6$s criterion for detection that allows for a maximum Type I error rate of <i>\u03B1=</i>%7$s.",
n_text, power, "\u03C1", sign, d, tail_text, alpha
)
} else if (calc == "effectSize") {
str <- gettextf(
"A design with %1$s will reliably (with probability greater than %2$s) detect variance ratios of <i>%3$s%4$s</i>%5$s, assuming a %6$s criterion for detection that allows for a maximum Type I error rate of <i>\u03B1=</i>%7$s.",
n_text, power, "\u03C1", sign, round(d, 3), tail_text, alpha
)
} else if (calc == "power") {
str <- gettextf(
"A design with %1$s can detect variance ratios of <i>%2$s%3$s</i>%4$s with a probability of at least %5$s, assuming a %6$s criterion for detection that allows for a maximum Type I error rate of <i>\u03B1=</i>%7$s.",
n_text, "\u03C1", sign, round(d, 3), round(power, 3), tail_text, alpha
)
}
hypo_text <- ifelse(alt == "two.sided",
gettextf("<i>%1$s</i>%2$s1", "\u03C1", "\u2260"),
ifelse(alt == "less", "<i>\u03C1<1</i>", "<i>\u03C1>1</i>")
)
str <- paste0(
str,
gettextf(
"<p>To evaluate the design specified in the table, we can consider how sensitive it is to true effects of increasing sizes; that is, are we likely to correctly conclude that %1$s when the variance ratio is large enough to care about?",
hypo_text
)
)
html[["text"]] <- str
table <- jaspResults[["powerEStab"]]
probs <- c(.5, .8, .95)
probs_es <- try(sapply(probs, function(p) {
if (alt == "two.sided" && d < 1) {
.pwrVarTest(
n = n, sig.level = alpha, power = p,
alternative = alt
)$rho[2]
} else {
.pwrVarTest(
n = n, sig.level = alpha, power = p,
alternative = alt
)$rho[1]
}
}))
if (inherits(probs_es, "try-error")) {
table$setError(gettext("The specified design leads to (an) unsolvable equation(s) while computing the values for this power table. Try to enter less extreme values for the parameters."))
return()
}
sign1 <- ifelse(alt == "less" || (alt == "two.sided" && d < 1), ">", "<")
sign2 <- ifelse(alt == "less" || (alt == "two.sided" && d < 1), "\u2265", "\u2264")
sign3 <- ifelse(alt == "less" || (alt == "two.sided" && d < 1), "\u2264", "\u2265")
esText <- c(
gettextf("1 %1$s %2$s %3$s %4$s", sign1, "\u03C1", sign2, format(round(probs_es[1], 3), nsmall = 3)),
gettextf("%1$s %2$s %3$s %4$s %5$s", format(round(probs_es[1], 3), nsmall = 3), sign1, "\u03C1", sign2, format(round(probs_es[2], 3), nsmall = 3)),
gettextf("%1$s %2$s %3$s %4$s %5$s", format(round(probs_es[2], 3), nsmall = 3), sign1, "\u03C1", sign2, format(round(probs_es[3], 3), nsmall = 3)),
gettextf("%1$s %2$s %3$s", "\u03C1", sign3, format(round(probs_es[3], 3), nsmall = 3))
)
cols <- list("es" = esText)
table$addColumns(cols)
if (alt == "two.sided" && calc != "effectSize") {
table$addFootnote(gettext("To improve readability of this power table for a 'Two-sided' alternative hypothesis, only the intervals in the direction of the specified variance ratio are displayed under 'True variance ratio'"))
}
}
#### Plot functions ----
.preparePowerContourTest1Var <- function(jaspResults, options, r, lst) {
image <- jaspResults[["powerContour"]]
if (is.null(image)) {
image <- createJaspPlot(title = gettext("Power Contour"), width = 400, height = 350)
image$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio",
"powerContour"
))
image$position <- 5
jaspResults[["powerContour"]] <- image
}
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
power <- ifelse(calc == "power",
r$power,
ifelse(calc == "sampleSize",
.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power,
lst$pow
)
)
ps <- .pwrPlotDefaultSettings
maxn <- try(ceiling(.pwrVarTest(
rho = d, power = max(0.99, power), sig.level = alpha, alternative = alt
)$n))
if (inherits(maxn, "try-error")) {
image$setError(gettext("The specified design leads to (an) unsolvable equation(s) while constructing the Power Contour plot. Try to enter less extreme values for the parameters"))
return()
}
if (n >= maxn && n >= ps$maxn) {
maxn <- ceiling(n * ps$max.scale)
} else if (maxn < ps$maxn) {
if ((ps$maxn - n) < 20) {
maxn <- ps$maxn * ps$max.scale
} else {
maxn <- ps$maxn
}
}
minn <- 2
try <- try(.pwrVarTest(n = minn, sig.level = alpha, power = power, alternative = alt))
while (inherits(try, "try-error")) {
minn <- minn + 1
try <- try(.pwrVarTest(n = minn, sig.level = alpha, power = power, alternative = alt))
}
nn <- unique(ceiling(exp(seq(log(minn), log(maxn), len = ps$lens)) - .001))
mind <- ifelse(alt == "two.sided" || alt == "less", min(d, 0.1), 1)
maxd <- ifelse(alt == "less", 1,
ifelse(alt == "two.sided" && d < 1, max(2, ((1 / d) * 1.2)), max(2, (d * 1.2)))
)
dd <- seq(mind, maxd, len = 20)
z.pwr <- try(sapply(dd, function(delta) {
.pwrVarTest(n = nn, rho = delta, sig.level = alpha, alternative = alt)$power
}))
if (inherits(z.pwr, "try-error")) {
image$setError(gettext("The specified design leads to (an) unsolvable equation(s) while constructing the Power Contour plot. Try to enter less extreme values for the parameters"))
return()
}
z.delta <- try(sapply(nn, function(N) {
.pwrVarTest(n = N, sig.level = alpha, power = power, alternative = alt)$rho
}))
if (inherits(z.pwr, "try-error")) {
image$setError(gettext("The specified design leads to (an) unsolvable equation(s) while constructing the Power Contour plot. Try to enter less extreme values for the parameters"))
return()
}
state <- list(
z.pwr = z.pwr,
z.delta = z.delta,
ps = ps,
nn = nn,
dd = dd,
n = n,
delta = d,
alpha = alpha,
minn = minn,
maxn = maxn
)
image$plotObject <- .plotPowerContour(options, state = state, ggtheme = .pwrPlotTheme())
}
.preparePowerCurveESTest1Var <- function(jaspResults, options, r, lst) {
image <- jaspResults[["powerCurveES"]]
if (is.null(image)) {
image <- createJaspPlot(
title = gettext("Power Curve by variance ratio"),
width = 400,
height = 350
)
image$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio",
"powerByEffectSize"
))
image$position <- 7
jaspResults[["powerCurveES"]] <- image
}
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
power <- ifelse(calc == "power",
r$power,
ifelse(calc == "sampleSize",
.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power,
lst$pow
)
)
ps <- .pwrPlotDefaultSettings
maxd <- try(.pwrVarTest(n = n, power = max(0.999, power), sig.level = alpha, alternative = alt)$rho)
if (inherits(maxd, "try-error")) {
mind <- ifelse(d < 1, d, 1)
maxd <- ifelse(d < 1, 1, d)
} else {
if (alt == "two.sided") {
mind <- min(as.numeric(maxd[2]), d)
maxd <- max(as.numeric(maxd[1]), d)
} else if (alt == "less") {
mind <- min(as.numeric(maxd), d)
maxd <- 1
} else {
mind <- 1
maxd <- max(as.numeric(maxd), d)
}
}
dd <- seq(mind, maxd, len = ps$curve.n)
y <- try(.pwrVarTest(n = n, rho = dd, sig.level = alpha, alternative = alt)$power)
if (inherits(y, "try-error")) {
image$setError(gettext("The specified design leads to (an) unsolvable equation(s) while constructing the power curve. Try to enter less extreme values for the parameters"))
return()
}
cols <- ps$pal(ps$pow.n.levels)
yrect <- seq(0, 1, 1 / ps$pow.n.levels)
state <- list(cols = cols, dd = dd, y = y, yrect = yrect, n = n, alpha = alpha, delta = d, pow = power)
image$plotObject <- .plotPowerCurveES(options, state = state, ggtheme = .pwrPlotTheme())
}
.preparePowerCurveNTest1Var <- function(jaspResults, options, r, lst) {
image <- jaspResults[["powerCurveN"]]
if (is.null(image)) {
image <- createJaspPlot(
title = gettext("Power Curve by N"),
width = 400,
height = 350
)
image$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio",
"powerBySampleSize"
))
image$position <- 9
jaspResults[["powerCurveN"]] <- image
}
calc <- options$calculation
ps <- .pwrPlotDefaultSettings
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
power <- ifelse(calc == "power",
r$power,
ifelse(calc == "sampleSize",
.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power,
lst$pow
)
)
maxn <- try(ceiling(.pwrVarTest(
rho = d, power = max(0.99999, power), sig.level = alpha, alternative = alt
)$n))
if (inherits(maxn, "try-error")) {
image$setError(gettext("The specified design leads to (an) unsolvable equation(s) while constructing the 'Power Curve by N' plot. Try to enter less extreme values for the parameters"))
return()
} else if (n >= maxn && n >= ps$maxn) {
maxn <- ceiling(n * ps$max.scale)
}
minn <- 2
try <- try(.pwrVarTest(n = minn, sig.level = alpha, power = power, alternative = alt))
while (inherits(try, "try-error")) {
minn <- minn + 1
try <- try(.pwrVarTest(n = minn, sig.level = alpha, power = power, alternative = alt))
}
nn <- seq(minn, maxn)
y <- try(.pwrVarTest(n = nn, rho = d, sig.level = alpha, alternative = alt)$power)
if (inherits(y, "try-error")) {
image$setError(gettext("The specified design leads to (an) unsolvable equation(s) while constructing the 'Power Curve by N' plot. Try to enter less extreme values for the parameters"))
return()
}
cols <- ps$pal(ps$pow.n.levels)
yrect <- seq(0, 1, 1 / ps$pow.n.levels)
lims <- data.frame(
xlim = c(minn, maxn),
ylim = c(0, 1)
)
state <- list(
n = n,
cols = cols,
nn = nn,
y = y,
yrect = yrect,
lims = lims,
delta = d,
alpha = alpha,
pow = power
)
image$plotObject <- .plotPowerCurveN(options, state = state, ggtheme = .pwrPlotTheme())
}
.preparePowerDistTest1Var <- function(jaspResults, options, r, lst) {
image <- jaspResults[["powerDist"]]
if (is.null(image)) {
image <- createJaspPlot(
title = gettext("Power Demonstration"),
width = 400,
height = 300
)
image$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio",
"powerDemonstration"
))
image$position <- 11
jaspResults[["powerDist"]] <- image
}
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
power <- ifelse(calc == "power", r$power, lst$pow)
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
df <- n
ncp <- n * d
if (alt == "two.sided") {
crit_upper <- qchisq(1 - alpha / 2, df = df) / n
crit_lower <- qchisq(alpha / 2, df = df) / n
} else if (alt == "greater") {
crit <- qchisq(p = 1 - alpha, df = df) / n
} else {
crit <- qchisq(alpha, df = df) / n
}
if (d > 1) {
xlims <- c(floor(qchisq(.0001, df)), qchisq(.9999, ncp)) / n
} else {
xlims <- c(floor(qchisq(.0001, ncp)), qchisq(.9999^(1 / d), df)) / n
}
if (d > 1) {
y.max <- dchisq(df - 2, df)
}
if (d < 1 && d > 0.3) {
y.max <- dchisq(ncp - 2, ncp)
}
if (d <= 0.3) {
y.max <- max(dchisq(ncp - 2, ncp), dchisq(df - 2, df) * 4)
}
xx <- seq(xlims[1], xlims[2], len = 1000)
yy.null <- dchisq(xx * n, df)
yy.alt <- dchisq(xx * n, ncp)
curves <- data.frame(
x = rep(xx, 2),
ymin = rep(0, length(xx) * 2),
ymax = c(yy.null, yy.alt),
group = rep(c("Null", "Alt"), each = length(xx))
)
if (alt == "two.sided") {
rect <- data.frame(
x1 = crit_lower, x2 = crit_upper,
y1 = 0, y2 = y.max * 1.1
)
} else if (alt == "greater") {
rect <- data.frame(
x1 = xlims[1] - 1, x2 = crit,
y1 = 0, y2 = y.max * 1.1
)
} else {
rect <- data.frame(
x1 = crit, x2 = xlims[2] + 1,
y1 = 0, y2 = y.max * 1.1
)
}
lims <- data.frame(
xlim = c(xlims[1], xlims[2]),
ylim = c(0, y.max * 1.1)
)
state <- list(curves = curves, rect = rect, lims = lims)
image$plotObject <- .plotPowerDist(options, state = state, ggtheme = .pwrPlotTheme())
}
#### Generate synthetic dataset ----
.generateDatasetTest1Var <- function(jaspResults, options, r, lst) {
datasetContainer <- jaspResults[["datasetcont"]]
if (is.null(datasetContainer)) {
# Create Container if it doesn't exist yet
datasetContainer <- createJaspContainer(title = gettext("Synthetic Dataset"))
datasetContainer$dependOn(c(
"test",
"effectSize",
"varianceRatio",
"effectDirection",
"power",
"sampleSize",
"alternative",
"alpha",
"calculation",
"sampleSizeRatio",
"savePath",
"firstGroupMean",
"secondGroupMean",
"firstGroupSd",
"secondGroupSd",
"populationSd",
"effectDirectionSyntheticDataset",
"testValue",
"setSeed",
"seed"
))
datasetContainer$position <- 12
jaspResults[["datasetcont"]] <- datasetContainer
generatedDataset <- createJaspState()
characteristicsTable <- createJaspTable(title = gettext("Characteristics"))
powerTable <- createJaspTable(gettext("Post Hoc Power Analysis"))
} else {
return()
}
# Generate dataset
if (!grepl(".csv", options[["savePath"]], fixed = TRUE) && !grepl(".txt", options[["savePath"]], fixed = TRUE)) {
.quitAnalysis(gettext("The generated dataset must be saved as a .csv or .txt file."))
}
calc <- options$calculation
n <- ifelse(calc == "sampleSize", r$n, lst$n)
if (calc == "effectSize") d <- ifelse(options$alternative == "twoSided", ifelse(options$effectDirection == "greater", r$es[1], r$es[2]), r$es) else d <- lst$es
alpha <- ifelse(calc == "alpha", r$alpha, lst$alpha)
alt <- lst$alt
power <- ifelse(calc == "power",
r$power,
ifelse(calc == "sampleSize",
.pwrVarTest(n = n, rho = d, sig.level = alpha, alternative = alt)$power,
lst$pow
)
)
sd_0 <- options[["populationSd"]]
var_0 <- sd_0^2
var_1 <- var_0 * d
sd_1 <- sqrt(var_1)
mean_1 <- options[["firstGroupMean"]]
if (options[["setSeed"]]) {
set.seed(options[["seed"]])
}
group_1 <- rnorm(n, mean = 0, sd = sd_1)
group_1 <- group_1 - mean(group_1)
group_1 <- group_1 * (sd_1 / sd(group_1))
group_1 <- group_1 + mean_1
id <- seq.int(1, n)
dependent <- group_1
dataset <- data.frame(cbind(id, dependent))
csv <- try(write.csv(dataset, options[["savePath"]], row.names = FALSE))
if (inherits(csv, "try-error")) {
.quitAnalysis(gettext("The generated dataset could not be saved. Please make sure that the specified path exists and the specified csv file is closed."))
}
generatedDataset <- dataset
datasetContainer[["generatedData"]] <- generatedDataset
# Characteristics tab
colNames <- c("n", "mean", "s", "s0")
colLabels <- c(
gettext("N"),
gettext("\u0078\u0305"),
gettext("s"),
gettext("\u03C3\u2080")
)
colType <- c("integer", "number", "number", "number")
for (i in seq_along(colNames)) {
characteristicsTable$addColumnInfo(colNames[i],
title = colLabels[i],
type = colType[i]
)
}
characteristicsTable[["n"]] <- n
characteristicsTable[["mean"]] <- mean_1
characteristicsTable[["s"]] <- sd_1
characteristicsTable[["s0"]] <- sd_0
characteristicsTable$addFootnote(gettextf("The synthetic dataset is saved as %s", options[["savePath"]]))
datasetContainer[["characteristics"]] <- characteristicsTable
# Post hoc power tab
colNames <- c("es", "alt", "power", "alpha")
colLabels <- c(
gettext("Variance ratio (\u03C1)"),
"Alternative hypothesis",
gettext("Power"),
"\u03B1"
)
colType <- c("number", "string", "number", "number")
for (i in seq_along(colNames)) {
powerTable$addColumnInfo(colNames[i],
title = colLabels[i],
type = colType[i]
)
}
powerTable[["es"]] <- d
powerTable[["alt"]] <- switch(alt,
"two.sided" = "Two-sided",
"less" = "Less (One-sided)",
"greater" = "Greater (One-sided)"
)
powerTable[["power"]] <- power
powerTable[["alpha"]] <- alpha
datasetContainer[["posthocpower"]] <- powerTable
}
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.