docsteveharris

# calculate average number of drug doses per patient
col.names.drugs <- names(Filter(function(y) (!is.null(y) && y=="Drugs") , map(dict, "Classification1")))
tdt <- copy(data2d.wide)
tdt[, id:= paste0(site, "_", episode_id)]
setkey(tdt, id)
dim(tdt)
names(tdt)
col.names <- col.names.drugs[col.names.drugs %in% names(tdt)]
tdt[1:10,print(.SD), by=id, .SDcols=col.names]
drug.doses <- tdt[,lapply(.SD, function(x) sum(ifelse(x=="NA",as.integer(NA),1L), na.rm=TRUE)), by=.(site, id), .SDcols=col.names]

docsteveharris

#' @title Mosaic plot using ggplot.
#'
#' @description
#' Creates a mosaic plot where the dimensions of the cells of a 
#' confusion matrix represent their marginal proportions.
#'
#' @details
#' Credit for initial iteration to 
#' [Edwin](http://stackoverflow.com/a/19258045/992999)
#' This version adds color brewer options and tidies the labelling

docsteveharris

# via https://stat.ethz.ch/pipermail/r-sig-mixed-models/2008q2/000874.html
# Calculate the median odds ratio
# -------------------------------
MOR <- function(my.var, digits = 2)
          { # MOR arguments: my.var = variance associated with level 2 clustering variable
            # digits = number of decimal places to which MOR value will be rounded.

            Median.OR <- round(exp(sqrt(2*my.var)*qnorm(.75)), digits)
            paste("Median Odds-Ratio (MOR) = ", Median.OR)  }

docsteveharris

/*

 ## Create a rug plot
_____________________
CreatedBy:	Steve Harris

CreatedAt: 	111291
ModifiedAt:	120701

Log

docsteveharris

gen.sofa.c <- function(bpsys, bpdia, rxcvs_drug=NULL, rxcvs_dose=NULL) {

    rx1 <- c("Adrenaline", "Noradrenaline")
    rx2 <- c("Dopamine")
    rx3 <- c("Vasopressin")
    rx  <- c(rx1, rx2, rx3, "Other")

    # Print these descriptions so you know what you have passed to the function
    bpmap <- round(bpdia + (bpsys - bpdia) / 3)

docsteveharris

score.charlson <- function(data) {
    # data is a datatable containing the columns you need
    # identify the columns with pmh in the name
    require(data.table)
    charlson.values <- data[,.(pmhmi, pmhhf, pmhpvd, pmhcvd, pmhdem, pmhcopd, pmhcopd, pmhctd, pmhud, pmhmld, pmhdm, pmhhemi, pmhckd, pmhdmend, pmhtumour, pmhleuk, pmhlymph, pmhsld, pmhmets, pmhaids)]
    charlson.matrix <- as.matrix(charlson.values)
    # Now I need to convert the NA's to 0 and >0 to 1 in matrix
    charlson.matrix <- apply(charlson.matrix, 2, function(x) ifelse(is.na(x), 0, ifelse(x>=1,1, 0)))
    # Now create a vector of scores
    # Note that the order of the columns selected above *must* match the order of the scores here

docsteveharris

# Test data: 1000 patients with two different age distributions
tdt <- data.table(age.old=rnorm(1000, mean=65, sd=15), age.young=rnorm(1000,mean=55,sd=10))

# Function to compare two distributions using qplot
qqplot <- function(x,y, data=wdt, n=100) {
    lab.x = x
    lab.y = y
    x <- with(data, get(x))
    x <- sapply(seq(0,1,1/n), function(q) quantile(x, q, na.rm=TRUE))
    y <- with(data, get(y))

docsteveharris

time24_regex = re.compile(
r"""\b
([01]?\d|2(?=[0-3])\d)
[:|.]?
([0-5]?\d)
\b""", re.VERBOSE)

docsteveharris

var dataset; // declare a global var to hold data
d3.csv("data/univar.csv", function(error, data) {

    if(error) {
        console.log(error);
    }
    else {
        console.log(data);
    }

docsteveharris

#! /usr/bin/python

# author: Steve Harris
# Convert short nvAlt image links to full path for previewing in Marked

# Todo
# ====
# - move the pathto line to a variable
# - post this script as a gist