plot_contour: Contour plot for subgroup effect size

Description

this function produces a contour plot showing the treatment effect size of subgroups. The subgroups are first defined by certain ranges of the first continuous covariate; and then further divided into smaller subgroup by certain ranges of the second covariate . The subgroups over the first covariate have a sample size close to one pre-specified value (N2) and any neighboring subgroups have an overlap size near the second pre-specified value (N1). Similarly, each subgroup over the first covariate has a sample size near the third pre-specified value (N4), and any neighboring subgroups which are further divided over the second covariate have a sample size near the fourth pre-specified value (N3). The x-coordinate and y-coordinate of a point indicates the middle point of the range over the first covariate and that over the second covariate, respectively. The contours show approximate effect sizes which are obtained by fitting grid points over the polynormial surface interpolating the points corresponding to subgroups.Note that there are three parameters for controlling the setting of contours. In addition, the function uses log odd ratio and log hazard ratio for displaying subgroup effect sizes in binary and survival data, respectively. Also, the actual subgroup sample sizes over the covariates are shown on the console window.

Usage

plot_contour(dat, covari.sel, trt.sel, resp.sel, outcome.type, setup.ss,
  n.grid = c(41, 41), brk.es = c(0, 1, 2, 3), n.brk.axis = 7,
  para.plot = c(0.35, 2, 20), font.size = c(1.5, 1.2, 1, 0.85, 0.8),
  title = NULL, subtitle = NULL, effect = "HR", point.size = 1.2,
  filled = FALSE, strip = NULL, show.overall = FALSE,
  palette = "divergent", col.power = 0.5, show.points = FALSE)

Arguments

dat

a data set

covari.sel

a vector of indices of the two covariates

trt.sel

a variable index specifying the treatment code

resp.sel

a variable index specifying the response variable

outcome.type

a string specifying the type of the response variable, it can be "continuous", or "binary" or "survival".

setup.ss

a vector specifying approximate subgroup sample size and neibourghing subgroup overlap sample size. The first and the second elements are for overlap sizes and subgroup sample sizes over the first covariate; the third and thefourth are for further divided overlap sizes and subgroup sample sizes over the second covariate.

n.grid

a vector specifying the numbers of the grid points on the x-axis and the y-axis respectively.

brk.es

a vector specifying the break points on effect size, where each range partition is given with a different colour on points.

n.brk.axis

a number specifying the number of breakpoints dividing the axis of the argument "range.strip".

para.plot

a vector specifying the parameters of the contour plot; the first value is for controlling the degree of smoothing; the second is for controlling the degree of the polynomials fitting to be used (normally 1 or 2); the third is for controlling the number of contour lines.

font.size

a vector specifying the size of labels and text; the first element is for the main title, the second is for for x-axis and y-axis labels; the third is for the subtitle; the fourth is for the text in the legend; the fifth is for the labels on contour lines.

title

a string specifying the main title.

subtitle

strings specifying the subtitle

effect

either "HR" or "RMST". only when outcome.type = "survival"

point.size

size of points for raw data points

filled

a logical indicator whether to show filled contour plot. If FALSE, contour lines are drawn

strip

the title for the strip showing treatment effect size

show.overall

logical. whether to show or not the overall treatment effect in the strip

palette

either "divergent" or "hcl"

col.power

to be used when palette = "hcl". see colorspace package for reference

show.points

a logical indicator specifying whether to show the raw data points

Examples

Run this code

# NOT RUN {
library(dplyr)

# Load the data to be used
data(prca)
dat <- prca
setup.ss =  c(10,60,15,30)
sub.title = bquote(N[11] %~~% .(setup.ss[2]) ~", "~
                     N[12] %~~% .(setup.ss[1]) ~", "~
                     N[21] %~~% .(setup.ss[4]) ~", "~
                     N[22] %~~% .(setup.ss[3]))
dat %>%
  rename(Weight = weight,
         Age = age) -> dat

plot_contour(dat,
             covari.sel = c(8,9),
             trt.sel = 3,
             resp.sel = c(1,2),
             outcome.type = "survival",
             setup.ss =  setup.ss,
             n.grid = c(100,100),
             brk.es = seq(-4.5,4.5,length.out = 101),
             n.brk.axis =  7,
             para.plot = c(0.5, 2, 6),
             font.size = c(1, 1, 1, 1, 1),
             title = NULL,
             subtitle = sub.title,
             strip = paste("Treatment effect size (log hazard ratio)"),
             show.overall = TRUE,show.points = TRUE,
             filled = TRUE, palette = "hcl",col.power = 0.75)

# }

Run the code above in your browser using DataCamp Workspace