martinmodrak

p <- MCMCpack::rdirichlet(1, rep(1, 4))[1,]

N <- 10000
g2 <- matrix(rnorm(length(p) * N), nrow = length(p))^2

pg2 = sweep(g2, MARGIN = 1, STATS = p, FUN = "*")
pg2_norm <- sweep(pg2, MARGIN = 2, STATS = colSums(pg2), FUN = "/")

martinmodrak

library(SBC)

library(cmdstanr)

library(bayesplot)
library(posterior)

library(future)
plan(multisession) 

martinmodrak

```{r}
library(cmdstanr)
```

```{r}
intercept_uncentered <- cmdstan_model(write_stan_file("data {
  int<lower=1> N;
  vector[N] signal;
  int<lower = 1> N_subj;
  vector[N] c_cloze;

martinmodrak

library(ggplot2)
N_sims <- 1000
N_draws <- 4000
sims <- matrix(rnorm(N_sims * N_draws), nrow = N_sims, ncol = N_draws)

q975 <- apply(sims, MARGIN = 1, FUN = quantile, prob = 0.975)
true_q975 <- qnorm(0.975)

ggplot(data.frame(x = q975), aes(x)) + 
  geom_histogram() +

martinmodrak

#Profile
ggplot(data.frame(x = 1:3, y = c(5, 7, 1)), aes(x, y)) + 
   geom_bar(stat = "identity", fill = "#6364ff", color = FALSE) + 
   theme_void() + 
   theme(plot.margin = margin(15,15,15,15))

# Header
set.seed(55465)
x <- 0:100
m <- splines::bs(x = x, df = 5, intercept = TRUE) %*% matrix(rnorm(5 * 50), ncol = 50, nrow = 5)

martinmodrak

library(ggplot2)
library(effsize)

set.seed(156226)
N <- 50
groups <- rep(c("A","B"), each = N)
means <-  rep(c(20,18), each = N)
noise <- rnorm(2*N, 0, 1)
noise <- noise - mean(noise)
df1 <- data.frame(group = groups, y = means + noise * 0.001)

martinmodrak

library(tidyverse)
library(lubridate)
quakes_raw <- read_csv("https://earthquake.usgs.gov/earthquakes/feed/v1.0/summary/all_month.csv")

quakes <- quakes_raw %>% mutate(hour = floor_date(time, "hour"),
                                day = floor_date(time, "day"))

quake_fanos <- quakes %>% group_by(hour, day, net) %>%
  summarise(count = n()) %>%
  group_by(day, net) %>%

martinmodrak

library(ggplot2)

N_treatment <- 709
N_control <- 699
N_sims <- 1e4
prob_event <- 0.1

# Assuming prob(event|treatment) = prob(event|control) 
# The case prob(event|treatment) > prob(event|control) produces fewer false positives (type-I errors)
pos_control <- rbinom(N_sims, size = N_control, prob = prob_event)

martinmodrak

```{r}
library(tidyverse)
library(gganimate)
theme_set(cowplot::theme_cowplot())
```


```{r}
quad <- function(x) {
  -x^2

martinmodrak

library(cmdstanr)

model_code <- "

data {
  int N;
  vector[N] y;
  int K;
  int<lower=1, upper=K> groups[N];
}