Package 'shorm'

Implement SHARP test repeatedly on a dose-response curve

Description

The SHARP shape detection test relies on a underlying random generation process, which may lead to ambiguous test results. To examin its uncertainty and obtain a more robust conclusion, we may with to repet the test on the same data.

Usage

RepeatSHARP(df, nRep, mixed = FALSE, xName, yName, rName, niter = 1000)

Arguments

df	Dose-response curve data in a table format. Dose and responses should be two separate columns with numeric values.
nRep	Number of repetitions of SHAPR test. Integer.
mixed	Logical indicator (TRUE for FALSE) for whether or not to use the mixed-model-based test.
xName	The column name for dose. Character string.
yName	The column name for response. Character string.
rName	The column name for random effect. Only used if mixed = TRUE. Character string.git a
niter	An integer for the number of iterations in SHARP test procedure

Value

A dataframe of test results from all repititions. Each row is one single test with four p values after Holm adjustment.

Examples

# Simulate dose-response data
x <- seq(0, 1, length.out = 48)
y <- 2*sqrt(x)+rnorm(48)
y[17:32] <- y[17:32]+0.5
y[33:48] <- y[33:48]+1
curve <- data.frame(x, y)
curve$rep <- rep(1:3, each = 16)

# Fixed-model based test
RepeatSHARP(curve, nRep = 10, xName = "x", yName = "y", niter = 10)

# Mixed-model based test
RepeatSHARP(curve, nRep = 10, mixed = TRUE, xName = "x", yName = "y", rName = "rep", niter = 10)

---

Function to visualize SHARP test results

Description

Function to visualize SHARP test results

Usage

SharpScatter(
  pinc,
  pdec,
  pconc,
  pconv,
  alpha = 0.05,
  scale = "log",
  label = NULL,
  size_point = 2,
  size_label = 5,
  ...
)

Arguments

pinc, pdec, pconc, pconv	SHARP test p-values corresponding to increasing, decreasing, concave and convex
alpha	significance threshold. alpha=NULL corresponding to no significance threshold
scale	how axis should be scaled for better visual presentation. Choose between none, log and linear. Please see details for the transformation fomula.
label	labels of the point to visualize. label=NULL means no labeling.
size_point, size_label	the size of points and label. size_lable is used when label is not NULL.
...	additional parameters passed to ggplot

Details

The shape scatter plot represents the four p-values (corresponding to four shape types) of SHARP test on a 2D space. it can not only visualize the conclusions of the test, but also the confidence level of the test. The plot surface is divided into two parts along two perpendicular directions, corresponding to two opposite shapes. The coordinates of points are derived from p-values such that:

• Inconclusive shapes are placed in the center area within the significance thresholds (p > 0.05)

• Significant shapes are placed in the edge area out of the significance thresholds (p < 0.05)

• The closer points are to the significance thresholds, the more ambiguous shape conclusion is

Axis can be scaled for better visual presentation. When scale = "log", a log-based transformation is implemented as follows:

$f(x) = 1-\frac{log(1+a(1-|x|))}{log(1+a)}$

where $a=\frac{1-2\alpha}{\alpha^2}$.

When scale = "linear", a piece-wise linear transformation is implemented:

$f(x) = 1-0.5\frac{(1-|x|)}{\alpha}I(|x|>1-\alpha)+\frac{0.5|x|}{1-\alpha}I(|x| \leq \alpha)$

In both case, the significant and insignificant regions are also evenly divided in each test.

Value

The plotted ggplot object

Examples

# Simulate dose-response data
x <- seq(0, 1, length.out = 48)
y <- 2*sqrt(x)+rnorm(48)
y[17:32] <- y[17:32]+0.5
y[33:48] <- y[33:48]+1
curve <- data.frame(x, y)
curve$rep <- rep(1:3, each = 16)

# Fixed-model based test
sharpt <- SHARPtest(curve, xName = "x", yName = "y", niter = 100)

# Plot the graph
SharpScatter(sharpt[1], sharpt[2], sharpt[3], sharpt[4], niter = 100)

---

Implement SHARP test once on a dose-response curve

Description

Implement SHARP test once on a dose-response curve

Usage

SHARPtest(df, mixed = FALSE, xName, yName, rName, niter = 1000)

Arguments

df	Dose-response curve data in a table format. Dose and responses should be two separate columns with numeric values.
mixed	Logical indicator (TRUE for FALSE) for whether or not to use the mixed-model-based test.
xName	The column name for dose. Character string.
yName	The column name for response. Character string.
rName	The column name for random effect. Only used if mixed = TRUE. Character string.
niter	An integer for the number of iterations in SHARP test procedure.

Value

A vector of four p values after Holm adjustment, indicating the significance of four different shapes types.

Examples

# Simulate dose-response data
x <- seq(0, 1, length.out = 48)
y <- 2*sqrt(x)+rnorm(48)
y[17:32] <- y[17:32]+0.5
y[33:48] <- y[33:48]+1
curve <- data.frame(x, y)
curve$rep <- rep(1:3, each = 16)

# Fixed-model based test
SHARPtest(curve, xName = "x", yName = "y", niter = 100)

# Mixed-model based test
SHARPtest(curve, mixed = TRUE, xName = "x", yName = "y", rName = "rep", niter = 100)

---