Lecture 10 ggplot2 and other interlopers.Rmd

---
title: "Lecture 10 ggplot2 and other interlopers"
author: "David McAllister"
date: "31st October 2016"
output:
  slidy_presentation: default
---

## Objectives of session

* To get a flavour of ggplot2
* See SQL and BUGS as examples of R interacting with other software


## Idea behind ggplot2

- Grammar of graphics
  + completely describes all standard graphics eg scatterplot, line, bar-chart
    + mapping of aesthetics, eg x = time, y = rate, colour = sex
    + geoms - dots, lines, ribbons, 
- Tufte's advice on visual display

# Encourages good graph design which presents all data together with summaries

## Example
```{r, echo=FALSE}
library(reshape2)
df <- data.frame (id = 1:20, age = rep(1:10,2), sex = rep(c("male", "female"),each = 10))
df <- df[rep(1:nrow(df),5),]
df$t <- rep(0:4, each = 20)
df$value <- with (df, age* 0 + 1.2*t + age*t*0.8 + rnorm (length(age), mean = 0, sd = 2))
df <- dcast (df, id + age + sex ~ t, value.var = "value")
df.melt <- melt(df, id = c("id", "age", "sex"))
df.melt$time <- as.numeric (df.melt$variable)
df.melt$variable <- NULL
df.melt
```

## Code for plot
```{r}
library (ggplot2)
plot1 <- ggplot (df.melt, 
                aes(x = time, y = value, colour = age, group = id)) +
                geom_point() +
                geom_line () +
                facet_grid (.~sex) 
```

## Change over time
```{r}
plot1
```

## Why is it an interloper
- Funny use of +
- completely separate to base R graphics
- No generic functions
  + if want to use it, understand lists and extract what you want
- Very nice striking graphics

## SQL within R
RSQLITE, RMySQL, SQLDF
```{r,eval= FALSE}
omopcon <- dbConnect(MySQL(), user=Sys.getenv("USER"), password=Sys.getenv("DBPASS"),
                     dbname="OMOP_2011", host="localhost")

dutasteride <- dbGetQuery (omopcon, 
          "SELECT * FROM o_drug_era INNER JOIN o_concept_drugs
             ON o_drug_era.concept_id = o_concept_drugs.UID
             WHERE drugname RLIKE 'DUTASTERIDE|AVODART'")
dbDisconnect(omopcon)
```

## Passes strings
All of these packages pass strings to other programs, retrieve output as objects in R

## SQL
- SELECT statement workhorse
- Easy to understand (unlike database design etc)
![alt text](figures/sams.jpg)
- Careful about
    + dots
    + changing variable types (work with characters)
    + example using Dundee server
  
## R2JAGS, the model
```{r, eval = FALSE}
modelString = "
# JAGS model specification begins ...
model {
    # Likelihood:
    for ( i in 1:n ) {
        y[i] ~ dbern( theta )
    }
    # Prior distribution:
    theta ~ dbeta( 1, 1 )
# Sample from the distribution estimating the prevalence 
# of the exposure in controls
     control ~ dbeta(10,40)
# Calculate the differences in proportions, ratios for 
# exposure and odds ratio
     oddscase <- theta/(1-theta)
     oddscontrol <- control/(1-control)
}
"
# Write the modelString to a file, using R commands:
writeLines(modelString,con="model.txt")
```

## Execution and data
```{r, eval = FALSE}
dataList = list(
    n = 13 ,
    y = c( 1,1,1,1,1,1,1,1,1,1,1,1,1 ))
jags <- jags.model('model.txt',
                   data = dataList,
                   n.chains = 1,
                   n.adapt = 1000)
update(jags, 1000)
```

## rjags
As for SQL pass string to another program, JAGS, and get data back