mathzero

#' Function requires a PubMed API key. They are free and easily obtained
#' Instructions get an API key:
#' https://ncbiinsights.ncbi.nlm.nih.gov/2017/11/02/new-api-keys-for-the-e-utilities/
#' Register here: https://account.ncbi.nlm.nih.gov/


#' This function runs a PubMed query and returns a data frame of metadata for the results
#' Where the query returns >10,000 (which normally triggers an error from the API), 
#' the function automatically breaks down the search into chunks by date and returns a combined and de-duped data set

mathzero

# Main function to find the 'optimal' combination of rows and columns to drop in order to 
# maximise the number of remaining data points in a data set

# Search space rapidly becomes enormous as data size grows, so this random search solution will
# only ever be a rough approximation of the optimal data set. 

optimiseDataset <- function(data, max_iterations = 10000) {

  best_solution <- data
  dims=dim(data[complete.cases(data),])

mathzero

# install.packages("pubmedR")
# install.packages("rcrossref")
# devtools::install_github("ropensci/rAltmetric")

library(pubmedR)
library(rcrossref)
library(rAltmetric)
pacman::p_load(curl,readr, tidyverse,purrr) # load required packages

mathzero

install.packages("pubmedR")
install.packages("rcrossref")
devtools::install_github("ropensci/rAltmetric")
install.packages("tidyverse")
install.packages("janitor")

library(pubmedR)
library(rcrossref)
library(tidyverse)
library(janitor)

mathzero

altmetrics_new <-
  function(doi = NULL,
           apikey = NULL,
           ...) {
    
    base_url <- "https://api.altmetric.com/v1/"
    args <- list(key = apikey)
    request <-
      httr::GET(paste0(base_url, "doi/",doi))
    if(httr::status_code(request) == 404) {

mathzero

library(ggplot2)
library(dplyr)
library(tidyr)
library(faux)
library(randomForest)


# Set parameters ----------------------------------------------------------

mathzero

# xgboost permutation function

#' This function takes an XGBoost model and some X and y data 
#' (ideally this would be unseen holdout data but you could also use the training data)
#' and returns a data frame with an estimation of the contribution that each variable makes to the overall AUC
#' can take a long time to run with a large data set – nperm can be reduced to reduce compute time


PermuteImportXBG <- function(model, X, y, nperm = 100){
  predictors=model$feature_names

mathzero

#' This short function takes some input from whatever R script you are working in and creates an 
#' rmarkdown document and corresponding knitted HTML file on the fly
#' Basis for code is: https://stackoverflow.com/questions/60110904/how-to-generate-html-report-directly-from-r-script


my_markdown_rederer <- function(text, myList=myList) {
  
  # The file name for the rmd doc
  rmd_file_name <- "temp.Rmd"
  

mathzero

#### 1. Cross tab function 

# Simple cross-tab function, with %s 
# Creates a cross tab of categorical variables, with %s in brackets after the number
# You can specify the names of the levels within the variables if you want, or leave blank to use the existing names
# (Be sure to enter the level names in the right order if you do specify them)

crossTab <- function(dat = dfRes, rowvar, colvar, rowvar_levels = NULL,colvar_levels = NULL){
  tab <- addmargins(table(pull(dat,rowvar), pull(dat,colvar))) %>% as.data.frame.matrix()
  tab.prop <- round(100*prop.table(table(pull(dat,rowvar), pull(dat,colvar)),1),1) %>% as.data.frame.matrix()

mathzero

### slightly adapted from here to save png or pdf and specify resolution etc: https://stackoverflow.com/questions/43051525/how-to-draw-pheatmap-plot-to-screen-and-also-save-to-file

# function detects '.png' or '.pdf' in the declared filename and assigns that file type

### pheatmap save function
save_pheatmap <- function(x, filename, width=12, height=12){
  stopifnot(!missing(x))
  stopifnot(!missing(filename))
  if(grepl(".png",filename)){
    png(filename, width=width, height=height, units = "in", res=300)