Mattan S. Ben-Shachar

@mattansb.msbstats.info

3642 Followers 401 Following

Statistics lecturer | Freelance statistical consultant & research analyst | #rstats dev @easystats.github.io home.msbstats.info (He/Him)

library(ggplot2)
library(distributional)
library(ggdist)
library(ggfx)

update_geom_defaults("slab", aes(fill = NA, linewidth = 1, alpha = 0.7))

img <- jpeg::readJPEG("msid-44847844,width-96,height-65-499916619.jpg")

p <- ggplot() +
annotation_raster(img, xmin = -9.5, xmax = 9.5, ymin = -0.2, ymax = 1.07) +
geom_hline(yintercept = 0, color = "black", linewidth = 1) |>
with_outer_glow("white") +
stat_slab(aes(xdist = dist_cauchy(0, 1), color = "Cauchy"), n = 10001) |>
with_outer_glow("white") +
stat_slab(aes(xdist = dist_laplace(0, 1), color = "Laplace")) |>
with_outer_glow("white") +
stat_slab(aes(
xdist = dist_student_t(3, 0, 1),
color = "Student t (df = 3)"
)) |>
with_outer_glow("white") +
stat_slab(
aes(xdist = dist_normal(0, 1), color = "Std. Normal"),
p_limits = c(.000000001, .999999999)
) |>
with_outer_glow("white") +
coord_fixed(
ratio = (19 / 1.27) / (1024 / 683),
xlim = c(-8, 8),
ylim = c(-0.1, 1),
expand = c(0, 0)
) +
scale_color_viridis_d(
NULL,
limits = rev(c("Cauchy", "Student t (df = 3)", "Laplace", "Std. Normal"))
) +
labs(
title = "Which distribution best fits the Hallgrímskirkja church facade?",
x = expression(theta),
y = NULL
) +
theme_minimal() +
theme_sub_axis_left(text = element_blank(), ticks = element_blank())

ggsave(
"church.png",
p,
width = 11,
height = 7,
units = "cm",
dpi = 600,
scaling = 1 / 1.8
)

Tails from Sonic facepalming

library(ggplot2)

mtcars_radar <- mtcars |>
tibble::rowid_to_column("model") |>
tidyr::pivot_longer(
cols = c(mpg, hp, wt),
names_to = "metric",
values_to = "value"
) |>
dplyr::mutate(
value = scales::rescale(value, to = c(0, 100)),
.by = metric
)

ggplot(mtcars_radar, aes(metric, value, color = factor(cyl))) +
stat_summary(geom = "point") +
stat_summary(aes(group = factor(cyl)), geom = "line") +
expand_limits(y = c(0, 100)) +
coord_radial(
start = 3.95 * pi / 2,
end = 0.05 * pi / 2,
inner.radius = 0.8
)

Six quilted hexagons that look like the tidyverse hex sticker

All R²=.999

4 chairs at a traditional Seder table. No family this year as it's not safe to travel.

library(brms)

svar <- stanvar(
scode = '
// flat beta prior on the intercept
real p_mean = inv_logit(b[1]);
lprior += beta_lpdf(p_mean | 1, 1);

// tight normal prior for the difference in probability
real p_diff = inv_logit(b[1] + b[2]) - p_mean;
lprior += normal_lpdf(p_diff | 0, 0.2);

', block = "tparameters")

data("mtcars")

fit <- brm(
am ~ 0 + Intercept + factor(vs),
data = mtcars,
family = bernoulli("logit"),

stanvars = svar
)

A dendrogram of a decision tree with a plot on each node showing the split-variable on x.