In debruine/faux: Simulation for Factorial Designs

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  out.width = "100%"
)
ggplot2::theme_set(ggplot2::theme_bw())
set.seed(8675309)

library(faux)
library(ggplot2)

Plotting Designs

The functions check_design() and sim_design() will automatically create plots of your design (unless you set plot = FALSE) so you can check you set it up correctly. You can also use the plot_design() function to plot a saved design list.

One-factor designs

p1 <- check_design(plot = FALSE) %>% plot_design()
p2 <- check_design(2, plot = FALSE) %>% plot_design()
p3 <- check_design(3, plot = FALSE) %>% plot_design()
p4 <- check_design(4, plot = FALSE) %>% plot_design()
p5 <- check_design(5, plot = FALSE) %>% plot_design()
p6 <- check_design(6, plot = FALSE) %>% plot_design()

cowplot::plot_grid(p1, p2, p3, p4, p5, p6, nrow = 2)

Two-factor designs

p1 <- check_design(c(2,2), mu = 1:4, plot = FALSE) %>% plot_design()
p2 <- check_design(c(2,3), mu = 1:6, plot = FALSE) %>% plot_design()
p3 <- check_design(c(2,4), mu = 1:8, plot = FALSE) %>% plot_design()
p4 <- check_design(c(3,2), mu = 1:6, plot = FALSE) %>% plot_design()
p5 <- check_design(c(3,3), mu = 1:9, plot = FALSE) %>% plot_design()
p6 <- check_design(c(3,4), mu = 1:12, plot = FALSE) %>% plot_design()

cowplot::plot_grid(p1, p2, p3, p4, p5, p6, nrow = 3)

Three-factor designs

check_design(c(2,2,2), mu = 1:2^3)

Four-factor designs

check_design(c(2,2,2,2), mu = 1:2^4)

Five-factor designs

check_design(c(2,2,2,2,2), mu = 1:(2^5))

Six-factor designs

check_design(c(2,2,2,2,2,2), mu = 1:(2^6))

Plotting Data

You can plot data created with faux using plot_design(), too. It will return a ggplot with a violin-boxplot by default.

One-factor data

data <- sim_design(2, 2, n = 20, mu = 1:4, plot = FALSE)
plot_design(data)

plot_design(data, geoms = "violin")

plot_design(data, geoms = "box")

plot_design(data, geoms = "pointrangeSD")

plot_design(data, geoms = c("violin", "pointrangeSE"))

plot_design(data, geoms = c("violin", "jitter"))

Two-factor data

data <- sim_design(2, 2, mu = 1:4, plot = FALSE)
plot_design(data)

Palettes

You can change the default brewer palette from "Dark2" to any of the options (see ?ggplot2::scale_colour_brewer for the options).

data <- sim_design(5, n = 20, mu = 1:5, plot = FALSE)
cowplot::plot_grid(
  plot_design(data, palette = "Purples") + ggtitle("Purples"),
  plot_design(data, palette = "Pastel2") + ggtitle("Pastel2"),
  plot_design(data, palette = "Spectral") + ggtitle("Spectral"),
  nrow = 1
)

You can also add a non-brewer palette (or any other ggplot function). You will get a message about a duplicate scale, but you can suppress this with suppressMessages() or in the r chunk.

plot_design(data) + 
  scale_fill_viridis_d() +
  theme_classic() +
  xlab("") +
  ggtitle("Plot with viridis fill and classic theme")

Change order of factors

List the factors you want to show in the order of fill/colour, x-axis, facet row(s), facet column(s).

data <- sim_design(c(2,2,2), n = 50, mu = 1:8, sd = 16, plot = FALSE)
abc <- plot_design(data, "W1", "W2", "W3", geoms = "violin")
acb <- plot_design(data, "W1", "W3", "W2", geoms = "violin")
bac <- plot_design(data, "W2", "W1", "W3", geoms = "violin")
bca <- plot_design(data, "W2", "W3", "W1", geoms = "violin")
cab <- plot_design(data, "W3", "W1", "W2", geoms = "violin")
cba <- plot_design(data, "W3", "W2", "W1", geoms = "violin")

cowplot::plot_grid(abc, acb, bac, bca, cab, cba, nrow = 3, 
                   labels = c("123", "132", "213", "231", "312", "321"))

You can also plot a subset of the factors.

# main effects, no interactions
intercept <- 0
effect_code <- list(a = -.5, b = .5)
W1 <- lapply(effect_code, `*`, 2)
W2 <- lapply(effect_code, `*`, -1)
W3 <- lapply(effect_code, `*`, 0)

mu <- list(W1a_W2a_W3a = intercept + W1$a + W2$a + W3$a,
           W1b_W2a_W3a = intercept + W1$b + W2$a + W3$a,
           W1a_W2b_W3a = intercept + W1$a + W2$b + W3$a,
           W1b_W2b_W3a = intercept + W1$b + W2$b + W3$a,
           W1a_W2a_W3b = intercept + W1$a + W2$a + W3$b,
           W1b_W2a_W3b = intercept + W1$b + W2$a + W3$b,
           W1a_W2b_W3b = intercept + W1$a + W2$b + W3$b,
           W1b_W2b_W3b = intercept + W1$b + W2$b + W3$b)

data <- sim_design(c(2,2,2), n = 50, mu = mu, sd = 2, plot = FALSE)

# make plots
geoms <- c("pointrangeSD")
W1 <- plot_design(data, "W1", geoms = geoms) + ggtitle("W1")
W2 <- plot_design(data, "W2", geoms = geoms) + ggtitle("W2")
W3 <- plot_design(data, "W3", geoms = geoms) + ggtitle("W3")
W12 <- plot_design(data, "W1", "W2", geoms = geoms) + ggtitle("W1 & W2")
W13 <- plot_design(data, "W1", "W3", geoms = geoms) + ggtitle("W1 & W3")
W23 <- plot_design(data, "W2", "W3", geoms = geoms) + ggtitle("W2 & W3")

cowplot::plot_grid(W1, W2, W3, W12, W13, W23, nrow = 2)

Facet Labels

Set the labeller to "label_both" if you want facets to include the factor label.

within <- list(
  pet = c(cat = "Cats", dog = "Dogs"),
  time = c(am = "Morning", pm = "Night"),
  grp = c(ctl = "Control", exp = "Experimental"),
  dose = 1:5
)

factor_labels  <- list(pet = "Type of Pet",
                       time = "Time of Day",
                       grp = "Group",
                       dose = "Treatment Dose")

sim_design(within, vardesc = factor_labels,
           mu = 1:40, sd = 5, plot = FALSE) %>% 
  plot_design("dose", "pet", "time", "grp", labeller = "label_both", palette = "Spectral")

debruine/faux documentation built on Jan. 18, 2025, 2:29 a.m.