---
title: "Binomial Parameters"
author: "Russell Almond"
date: "September 1, 2020"
output: html_document
runtime: shiny
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(shiny)
library(ggplot2)
```
The [binomial distribution](https://en.wikipedia.org/wiki/Binomial_distribution) can be thought of as a number of draws, $n$, from
an urn with a proportion $p$, of black balls.
The probability of drawing exactly $x$ balls from an this urn is:
$$ p(X|n,p) = \binom{n}{X} p^X (1-p)^{n-X}$$
The expected value is $np$, and the standard deviation is
$\sqrt{np(1-p)}$.
Sometimes we write this in terms of the proportion of black balls in
the sample. That is $p$, with a standard deviation of $\sqrt{p(1-p)/n}$.
```{r density, echo=FALSE}
inputPanel(
sliderInput("n", label = "Number of draws:",
min=0, max=100, value=10, step=1),
sliderInput("p", label = "Probability of success:",
min = 0, max = 1, value = .6, step = 0.01)
)
renderPlot({
n <- as.numeric(input$n)
p <- as.numeric(input$p)
dat <- data.frame(x=0:n,y=dbinom(0:n,n,p))
ggplot(dat,aes(x,y)) +geom_col()
})
```
Note that this distribution is positively skewed if $p < 0.5$ and
negatively skewed if $p > 0.5$.
Note how when $n$ gets large, the binomial distribution looks a lot
like the normal. This is one of the first central limit theorems that
was discovered. (The closer that $p$ is to 0 or 1, the longer
convergence to the normal takes.)