dmarx

# Finding best cut-off for constructing a categorical variable 
# logistic regression

data(iris)

x0 = iris[iris$Species != 'setosa',]
plot(x0, col=x0$Species)

# Keep things simple for this demo
form = "is_virginica ~ Petal.Length + Petal.Width"

dmarx

# From the docs for mcglm::ahs

require(mcglm)
data(ahs, package="mcglm")
form1 <- Ndoc ~ income + age
form2 <- Nndoc ~ income + age

Z0 <- mc_id(ahs)
fit.ahs <- mcglm(linear_pred = c(form1, form2),
                 matrix_pred = list(Z0, Z0), link = c("log","log"),

dmarx

##########################################################
# Get data from the poisson demo at:                     #
#  https://stats.idre.ucla.edu/r/dae/poisson-regression/ #
##########################################################

p <- read.csv("https://stats.idre.ucla.edu/stat/data/poisson_sim.csv")
p <- within(p, {
  prog <- factor(prog, levels=1:3, labels=c("General", "Academic", 
                                            "Vocational"))
  id <- factor(id)

dmarx

DIRS := $(filter dir%, $(shell ls))

foo_sources := $(wildcard */source/foo.a)
foo_targets_prt := $(patsubst %.a, %.b, $(foo_sources))
foo_targets := $(subst source,target, $(foo_targets_prt))

bar_sources := $(wildcard */source/bar.a)
bar_x := $(patsubst %/bar.a, %/Y.a, $(bar_sources))
bar_y := $(patsubst %/bar.a, %/Z.a, $(bar_sources))
bar_targets :=  $(bar_x) $(bar_y)

dmarx

#' ---
#' title: "Regression for quantifying a regime change"
#' author: "David Marx"
#' date: "June 5, 2017"
#' output: html_document
#' ---

#' There are two time points of interest. We want to test the hypothesis that the regression 
#' coefficients changed after these time points, respectively. We will accomplish this by introducing
#' dummy variables to denote whether we are before or after a particular change point. This approach 

dmarx

To Read:

Publication Date	Article	Notes
2016	End-to-End Relation Extraction using LSTMs on Sequences and Tree Structures	Cited in multi-task sciERC (2018, below)
2018-10-11	BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
	Multi-Task Identification of Entities, Relations, and Coreference for Scientific Knowledge Graph Construction	Probably a lot of useful citations in here, not sure we need the coreference stuff. * SciERC datasets: http://nlp.cs.washington.edu/sciIE/ * Code: https://bitbucket.org/luanyi/scierc/src/master/ * Pretrained (best) models: NER, Coref, Relation
2017-08-08	[Structural

dmarx

# chinese restuarant process
chinese_restaurant = function(n, uniform=FALSE){
  tables = c(1) # running counts of people at tables. Start by seating first person at their own table
  U = runif(n)
  for (i in 2:n){
    if(U[i]<1/i){
      tables = c(tables, 1)
    } else {
      p = tables/(i) # sum(tables) = i-1
      

dmarx

generate_distances = function(k){
  u_k = c(0,sort(runif(k-1)),1)
  u_k[-1] - u_k[-(k+1)]
}

iters=1e4

d = c(replicate(iters, generate_distances(2)))
plot(density(d), ylim=c(0,5))
#abline(v=mean(d), lty=2)

dmarx

#' Extract the backbone of a weighted network using the disparity filter
#'
#' Given a weighted graph, \code{backbone} identifies the 'backbone structure'
#' of the graph, using the disparity filter algorithm by Serrano et al. (2009).
#' @param graph The input graph.
#' @param weights A numeric vector of edge weights, which defaults to
#' \code{E(graph)$weight}.
#' @param directed The directedness of the graph, which defaults to the result
#' of \code{\link[igraph]{is_directed}}.
#' @param alpha The significance level under which to preserve the edges, which

dmarx

#install.packages('venneuler')
library(venneuler)

venn_intersection_text = function(venn, classes, label, adjustment=0.5, xadj=0, yadj=0 ){
  # fits a line between the centers of two classes and draws label text at the midpoint of that line + adjustment
  xv = adjustment*venn$centers[classes[1],1] + (1-adjustment)*venn$centers[classes[2],1] + xadj
  yv = adjustment*venn$centers[classes[1],2] + (1-adjustment)*venn$centers[classes[2],2] + yadj
  text(x=xv, y=yv, labels=label)
}