Statistical Software --- STAT-F-413 --- Maarten Jansen and Toufik Zahaf

Objectives

• Retrieve and analyse your own real data
• Use at least two different software systems and two different types of analyses (typically ANOVA and regression, but others are equally welcome: principle component analysis etc.)
• Find your data
  1. at a company, hospital, banks, insurance company: this option is by far the best. If you get data, then also try to get to know what sort of business questions the company/organization is trying to answer: use the data to respond to the questions.
  2. Otherwise (but less preferable) on the internet, e.g.: government data (such as statbel.gov.be) This option has the drawback that it is harder to be original and harder to focus on specific business questions.
     The data should be original, in the sense that they must not be popular in scientific papers or textbooks as illustration of a method.
     • Number of births per communality
     • Macro-economical data; per country, european, regional, provinces etc.
     • Socio-economical data
     Not allowed:
     • Time series: time dependence of your data is allowed (longitudinal), but time must not be the dominant explanatory variable
     • Birth weights of babies

Deadlines

Available material

• Class schedule, time table

• Introduction and descriptive statistics

  Slides introduction
  Slides Descriptive Statistics

• ANOVA and SAS OnDemand for Academics: Enterprise Guide

  Follow this link at SAS to create a profile (user id) at the SAS website and again to register as student for this course. (Registration as student is by "buying" a license for 0 euro)

• The R project for Statistical computing

  • Slides of the class on R

    These slides include the following data files:
    • mesquite.txt (data file)
    • catdataexemple.txt (data file)

  • The R homepage

  • CRAN: the Comprehensive R Archive Network

    This page links to a list of packages that can be downloaded into your installation of R. Some of these packages come with documentation (i.e., an online book), which is listed in the link behind the Contributed documentation
    

    How to install a package on your computer?

    • Download and installation proceed at once from within R using the command
      
      install.packages("mypackage",dependencies = TRUE)
      
      Mind the quotes and mypackage is of course the package name selected from the list (case sensitive)
    • In Windows R opens by default as a Graphical User Interface (the R-Gui), which allows you to install packages by clicking on the packages-button in GUI-menu, which starts a dialogue
    • After installation of such an optional package, this package has to be loaded in every R-session (i.e.,every time you start R). This is done with
      
      library(mypackage)
      
      (No quotes this time). The package is installed in a directory under the library directory of the R installation.
    • After loading the package, you need to load the data that come with it. This is with the command
      
      data(mydataset)
      
      (where mydataset has to be replaced with the correct name of the data set that you want to use)

• A class on R

  We will implement in R a routine that performs a simulation study about the construction of confidence intervals for the mean of a normal random variable with unknown standard deviation. We will implement this using vectorization where possible in order to avoid unnecessary loops. We also plot the resulting confidence intervals, thereby visualizing the issue whether the CI contains the true parameter value.
  A raw version of the code