gibran hemani explodecomputer

https://www.bris.ac.uk/integrative-epidemiology/intranet/training/

Causal methods

Introduction

In an epidemiological context, whether stated explicitly or not, we typically are referring to a counterfactual condition. A counterfactual condition is where we suppose that the outcome would not have happened in an alternate reality where all things were identical except for the exposure.

For example, if an individual had body mass index = 23 instead of body mass index = 27, but all else was identical, what would the impact be on their bone mineral density?

	# Get the list of arguments passed to R
	args <- commandArgs(T)

	# Get the first argument, make it numeric because by default it is read as a string
	job_id <- as.numeric(args[1])

	# Print the argument
	message("job number ", job_id)

	n <- 5
	a <- combn(c(1:(2*n)), 2)
	b <- combn(1:ncol(a), n)
	l <- list()
	m <- list()
	j <- 1
	for(i in 1:ncol(b))
	{
	x <- a[, b[,i]]
	if(!any(duplicated(c(x))))

	# How does adjusted r-square do with overfitting?

	rm(list=ls())
	set.seed(101)

	library(ggplot2)
	library(tidyr)

	y <- rnorm(1000)
	x <- matrix(rnorm(1000 * 1000), 1000, 1000)

	dat_3a$rsq.exposure <- with(dat_3a, 2 * beta.exposure^2 * eaf.exposure * (1-eaf.exposure))
	dat_3a$rsq.outcome <- with(dat_3a, 2 * beta.outcome^2 * eaf.outcome * (1-eaf.outcome))
	dat_3a$steiger_dir <- dat_3a$rsq.exposure > dat_3a$rsq.outcome

	dat_3b$rsq.exposure <- with(dat_3b, 2 * beta.exposure^2 * eaf.exposure * (1-eaf.exposure))
	dat_3b$rsq.outcome <- with(dat_3b, 2 * beta.outcome^2 * eaf.outcome * (1-eaf.outcome))
	dat_3b$steiger_dir <- dat_3b$rsq.exposure > dat_3b$rsq.outcome

	---
	title: Calculating HWE
	---

	Choose a set of parameters. For each allele frequency and sample size, run 100 simulations

	```{r}
	library(tidyr)
	library(ggplot2)
	library(dplyr)

	library(devtools)
	install_github("explodecomputer/simulateGP")
	library(simulateGP)
	library(ggplot2)

	param <- rbind(
	expand.grid(
	ncase=100,
	ncontrol=100,
	prev=0.001,

	randomness <- function(x)
	{
	a <- (sum(x[-1] != x[-length(x)]) / (length(x)-1) - 0.5) * 2
	return(round(a, 2))
	}
	make_dat <- function()
	{
	dat <- read.table("~/Desktop/ascii.txt", he=F, sep="\t", stringsAsFactors = FALSE, quote="", comment="")
	names(dat) <- c("dec", "hex", "binary", "al")
	dat$binary <- gsub("0", ".", dat$binary)

	library(dplyr)
	library(MeSH.db)
	k <- keys(MeSH.db, keytype = "MESHID")

	select(MeSH.db, keys = k[1:10], columns = c("MESHID", "MESHTERM"), keytype = "MESHID")

	LEU <- select(MeSH.db, keys = "Leukemia", columns = c("MESHID", "MESHTERM", "CATEGORY", "SYNONYM"), keytype = "MESHTERM")


	library("MeSH.AOR.db")

	---
	title: The mystery of Nick's unpaired socks
	Authors: The Balfour Road consortium
	---

	## Introduction

	Nick has only 16 socks and none of them form a pair. There are several mechanisms that could give rise to his predicament. The most likely follows a scenario approximating the following.

	Nick buys a pair of socks, and after some period of time, he loses one of the socks. Nick then buys another pair of socks, and eventually loses one of those too. And so on. If the time to loss of a single sock is Poissonly distributed with $\lambda$ time to event, then the expected amount of time taken to reach a situation with $n$ unpaired socks is simply $n\lambda$.