gganimate_moderation_example.R

library(tidyverse)
library(emmeans)
library(gganimate)
library(tweenr)

# Setup data --------------------------------------------------------------

n <- 200
df <- tibble(X1 = rnorm(n,100,15),
             X2 = rnorm(n,30,5),
             Y = -30*X1 + X1*X2 + 20 + rnorm(n))

# Fit Model ---------------------------------------------------------------

fit <- lm(Y ~ X1*X2, df)
summary(fit)

emtrends(fit,~X2,'X1',cov.red = range)

# Get Data for the Frames -------------------------------------------------

# Line Data
reg_line <- emmeans(fit,~X1*X2,
                    at = list(X1 = range(df$X1),
                              X2 = seq(min(df$X2),max(df$X2), length.out = 5))) %>%
  summary() %>%
  split(.$X2) %>%
  tween_states(1,0, 'linear',50)

# Points Data
nframes <- length(unique(reg_line$.frame))
step_size <- 5*diff(range(reg_line$X2))/(nframes-1)

filter_range <- function(df,frame,m_name,step_size){
  value <- reg_line[[m_name]][reg_line$.frame==frame][1]
  subset(df,
         df[[m_name]] > value - step_size & df[[m_name]] < value + step_size)
}

df_plot <- df %>%
  list() %>%
  rep(nframes) %>%
  map2(seq_len(nframes),filter_range,m_name = 'X2', step_size = step_size) %>%
  map2_dfr(seq_len(nframes),~mutate(.x,.frame = .y))

# Looks good?
head(df_plot)
head(reg_line)


# Plot --------------------------------------------------------------------

# setup
p <- ggplot(mapping = aes(X1,color = X2)) +
  geom_point(aes(y = Y), data = df_plot, size = 1.5, alpha = .7) +
  geom_line(aes(y = emmean), data = reg_line, size = 1) +
  scale_color_viridis_c() +
  theme_minimal()

# Static plot
p_static <- p + facet_wrap(~.frame)

# Animate
p_anim <- p +
  transition_states(.frame) +
  shadow_mark(T,T, exclude_layer = 2,color = 'gray', size = 1)

animate(p_anim,nframes = nrow(reg_line), duration = 2)

The beautiful result: