session2

Mandy
30.05.2016

Recap

We learned about

  • data structures:
    • vectors
    • data frames
  • assigning
    • =
    • <-
  • Functions e.g.:
    • c()
    • plot()
    • nrow()
    • summary()

Exercise

  • use R to calculate \( 4^3 + 3^{2+1} \)
  • Let our small data set be: 2 5 4 10 8
    1. Enter these data into a data vector v
    2. Find the square of each number.
    3. Substract 6 from each number
    4. Substract 9 from each number and then square the answer
  • Accessing parts of data frames via
    • $
    • and numbers using square brackets [ ]
    • vectors: my.vector[index]
    • data frame: my.dataframe[rows.index,col.index]

Exercise

  • load the ChickWeight data set into the workspace using data()
  • how many observations and how many variables? How did you get the information?
  • what is the mean, min and max of weight and Time
  • use plot() to produce a boxplot of weight dependent on Diet

The working directory

  • in every R session you can define the so called working directory
  • the working directory is the place (folder) on your hard drive or usb stick where R is saving objects, or is looking for files you wanna load
  • you can check what the working directory by typing getwd()
  • you can set the working directory e.g.
    • using setwd(“path/to/the/course/session2”)
    • using the menu: Session -> Set Working Directory -> Choose Directory

The working directory

Example:

setwd("/media/mandy/Volume/transcend/life/2016kurs/session2")
  • R comes from unix systems - so even on Windows you have to use the slash - not the backslash
  • you can set the working directory via the menu - and copy the line to your script afterwards

Exercise

  • check your working directory
  • set the working directory to today's sesssion folder if not already done

Packages

  • you can see the list of packages installed already on your computer looking on the Packages pane in the lower right panel in RStudio
  • marked packages are already attached to you workspace (i.e. you can use the functionality from the package)
  • you can load a package by marking the package or by typing 
library(package.name)
  • for example 
library(MASS)
  • using the library() command is recommended for scripts

Install Packages

  • you can use search for installed packages
  • if a package is not already installed you can use the Install menus inside the Packages pane or via the menu: Tools -> Install Packages
  • begin typing the name(s) of the packages you wanna install
  • please be sure that the install dependencies is marked
  • click Install

Exercise

  • check if the packages ggplot2, dplyr, readxl are already installed
  • if not: install them
  • load them (add the respective commands to your session2 script)

Load data into R

R can import data of different forms, from different file types, etc. It is important to remember: for almost every kind of file type you have to use a different function, e.g.

File Type function to load the data package
.rdata (R's own format) load() base
.csv (English) read.csv() base
.csv (German) read.csv() base
.txt read.table() base
.xlsx read_excel() readxl
.sav (SPSS) spss.get() Hmisc
.dta (Stata) stata.get() Hmisc
.sasbdat (SAS) sas.get() Hmisc
.sasxport (SAS Transport Files) sasxport.get() Hmisc

Load data into R

  • the syntax is the same for any of these files:
    • use the function with the path of the file you wanna load as argument
    • sometimes there are further arguments to specify e.g. for an excel file which sheet you wanna to import
    • assign the result to a suitable name
    • do not type the whole path: use the auto completion
library(readxl)
x1 <- read_excel("data/20160523evaluation_Kopie.xlsx",1)

Exercise

  • with excel_sheets(“path/to/filename.xlsx”) you read out the names of existing worksheets; use the command on the excel file from above!
    • how many data sheets do the file have?
  • create three further dataframes x2, x3, x4 from the other three excel sheets

Exercise

x2 <- read_excel("data/20160523evaluation_Kopie.xlsx",2)
x3 <- read_excel("data/20160523evaluation_Kopie.xlsx",3)
x4 <- read_excel("data/20160523evaluation_Kopie.xlsx",4)

Exercise

  • create another data frame from the file mz2010_cf.sav (data folder). Hints:
    • what is the type of the file?
    • what command do you need to load such a file?
    • which package contains this command?
    • have you the package already installed?
    • did you load the package?
library(Hmisc)
yy <- spss.get("data/mz2010_cf.sav")

Exercise

  • the documention of this data file is also contained in the data folder
    • look for the variable for sex (geschlecht) - what is the resp. column name?
    • what information is contained in the variable EF310
    • use the command table() on the these two columns, e.g.
  table(yy$EF310)

ggplot2

  • ggplot2 is a package to create graphics for publications
  • the gg stands for grammar of graphics (Leland Wilkinson, second edition 2005)
  • the package is written by Hadley Wickham
  • there is a cheat sheet available via the menu: Help -> Cheatsheets (internet connection required)
  • extensive documentation: http://docs.ggplot2.org/
  • the definition of a ggplot graphic needs
    • to specify the data frame containing information
    • a mapping from the variables in the data frame to characteristic of the graphics (aes())
    • a specification of the used geometry (i.e. what kind of plot do you want)
  • it sounds a bit complicated, but the logic is powerful and makes it very easy to create complicated graphics

Example/Exercise

  • what do you think - to visualize the information contained in the tabulation of the EF310 variable what kind of graphics would be appropriate? (script)

GGPLOT2 Examples

library(ggplot2)
ggplot(yy, aes(x = EF310)) +
  geom_bar()

plot of chunk ggplot2ex1

ggplot(yy, aes(x = EF310, colour = EF46)) +
  geom_bar()

plot of chunk ggplot2ex2

ggplot(yy, aes(x = EF310, colour = EF46)) +
  geom_bar(position = "dodge")

plot of chunk ggplot2ex3

ggplot(yy, aes(x = EF310, colour = EF46)) +
  geom_bar(position = "fill")

plot of chunk ggplot2ex4

ggplot(yy, aes(x = EF310, fill = EF46)) +
  geom_bar(position = "fill")

plot of chunk ggplot2ex5

ggplot(yy, aes(x = EF310, fill = EF46)) +
  geom_bar(position = "fill") +
  coord_flip()

plot of chunk ggplot2ex6

library(stringr)
ggplot(yy, aes(x = EF310, fill = EF46)) +
  geom_bar(position = "fill") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 25)) + 
  coord_flip()

plot of chunk ggplot2ex8

library(scales)
ggplot(yy, aes(x = EF310, fill = EF46)) +
  geom_bar(position = "fill") +
  ylab("proportion") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 25)) + 
  scale_y_continuous(labels = percent) + 
  coord_flip()

plot of chunk ggplot2ex9

ggplot(yy, aes(x = EF310, fill = EF46)) +
  geom_bar(position = "fill") +
  ylab("proportion") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 25)) + 
  scale_y_continuous(labels = percent) +
  scale_fill_manual(values = c("männlich"="midnightblue",
                               "weiblich"="deeppink4")) +
  coord_flip()

plot of chunk ggplot2ex10 1. the variable E49 in the mz data set contains the information of the about the marital status

  • plot the distribution of this variable using a barplot
  • colour the bars according to sex
  • use fill to convert into proportions
    1. the variables EF20 and EF40 contain numeric information
  • use ggplot() to create a scatterplot EF44 on the x-axis and E20 on the y-axis (you need geom_point() instead of geom_bar())
  • colour the points according to colour…
  • interpret! Whats a possible problem

Graphics examples from Monday

ggplot(yy,aes(x = EF310, fill = EF46)) +
  geom_bar(position = position_dodge(), alpha = 0.4) +
  scale_x_discrete(labels = 1:8) +
  scale_fill_manual(values = c("männlich" = "midnightblue",
                               "weiblich" = "deeppink3")) + 
  ylab("Anzahl Personen") + 
  xlab("hoechster Schulabschluss") +
  ggtitle("Barplot")

plot of chunk unnamed-chunk-7

Explanations

  • the position = position_dodge() places the bars for different sexes next to each other
  • scale_x_discrete(labels = 1:8): here we set change the labels on the x-axis to the codes, we use the discrete specification because we have a discrete x-axis
  • alpha setting alpha at a value between 0 and 1 sets the transparency of the filling of the bars: 0 is completely transparent (i.e. invisible), 1 is completely opaque
  • scale_fill_manual() customize the colour scale for fillings: here we set manually “maennlich” to “midnightblue” and “weiblich” to “deeppink”
  • ylab(), xlab() and ggtitle() can be used to change the titles of the axes, resp. the main title. An alternative would be the usage of **labs(y = “y-axis”,x=“x-axis”,title = “title”)
library(ggthemes)
ggplot(yy,aes(x = EF310, fill = EF46)) +
  geom_bar(position = position_fill(),colour="black") +
  scale_x_discrete(labels = 1:8) +
##  scale_fill_grey() + 
  ylab("Anzahl Personen") + 
  xlab("hoechster Schulabschluss") +
  ggtitle("Barplot") +
  facet_wrap(~EF49,nrow = 2) +
  theme_stata()

plot of chunk unnamed-chunk-8

Explanations

  • the ggthemes package provides predefined graphics designs, so you can have e.g. excel or stata look-alike graphics
  • position = position_fill() stretch you bars along the whole y-axis and shows the proportions of the sexes (transformation to percentages)
  • facet_wrap() creates facets defined by a formula like syntax (we will here about formulas soon)

Next time

Levels of measurement

dplyr