jknowles

// Additional tools for machine learning and predictive analytics in stata
/*
Author: Jared Knowles
Date: 09/12/2018
Purpose: Survey of some additional code helpful in conducting and explaining
or demonstrating predictive analytics to stakeholders.

You do not need to run all of this code - this is a survey of commands that
tackle different techniques. Pick and choose what might be most useful to you.
*/

jknowles

// Use Gists to store code you would like to remember later on
console.log(window); // log the "window" object to the console

jknowles

# Using the dataUSA.io API for Census Data in R

This gist contains some notes on constructing a query for census and economic data from the [DataUSA.io](http://datausa.io/) site. This is a quick-start guide to their API; for in-depth documentation check out their [API documentation](https://github.com/DataUSA/datausa-api/wiki/Overview).

A great way to learn how to structure a query is to visit a specific datausa.io page and click on the "Options" button on top of any graph, then select "API" to see the query syntax that created the graph.


![Analytics](https://ga-beacon.appspot.com/UA-27835807-2/gist-id?pixel)

## Example Use

jknowles

predict.robust <- function(model, data, robust_vcov = NULL, level = 0.95, 
                           interval = "prediction"){
  # adapted from
  # https://stackoverflow.com/questions/38109501/how-does-predict-lm-compute-confidence-interval-and-prediction-interval
  # model is an lm object from r
  # data is the dataset to predict from
  # robust_vcov must be a robust vcov matrix created by V <- sandwich::vcovHC(model, ...)
  # level = the % of the confidence interval, default is 95%
  # interval = either "prediction" or "confidence" - prediction includes uncertainty about the model itself
  if(is.null(robust_vcov)){

jknowles

# Calculate the AUC of a GLM model easily
# Jared Knowles
# model = a fitted glm in R
# newdata = an optional data.frame of new fitted values
auc.glm <- function(model, newdata = NULL){
  if(missing(newdata)){
    resp <- model$y
    # if(class(resp) == "numeric"){
    #   resp <- factor(resp)
    # }

jknowles

#Set Environment Variables
TEXINPUTS="C:\\" #Path to tex file in Windows

Sys.setenv(TEXINPUTS="C:\\~", BIBINPUTS=TEXINPUTS,BSTINPUTS=TEXINPUTS) 

#Path to texfiles in Windows, set BIB files and BST files the same

#Run before clicking "Compile PDF"

jknowles

library(shiny)
library(scales)

shinyServer(function(input,output){
  trialInput<-reactive(function(){
    bias<-input$coin
    sims<-input$obs
    reps<-input$reps
    trials<-rbinom(reps,sims,0.5+bias)
  })

jknowles

# Script to demonstrate distributions
library(eeptools)
library(shiny)
library(ggplot2)


rnormcor <- function(x,rho) rnorm(1,rho*x,sqrt(1-rho^2))

shinyServer(function(input,output){
  output$distPlot<-reactivePlot(function(){

jknowles

# Script to demonstrate distributions
library(VGAM)
library(eeptools)
library(shiny)
library(ggplot2)




shinyServer(function(input,output){

jknowles

library(shiny)

shinyServer(function(input,output){
  
  output$distPlot<-reactivePlot(function(){
    dist<-rnorm(input$obs)
    p<-qplot(dist,binwidth=0.1)+geom_vline(xintercept=mean(dist))+theme_dpi()
    p<-p+coord_cartesian(xlim=c(-4,4))+geom_vline(xintercept=median(dist),color=I("red"))
    print(p)
  })