Workflow of Statistical Analysis with STATA: Data Management, Analysis and Visualization

Kommentar/Inhalt

In their education
students do regularly get to know statistics to a certain extent. Ideally,
some data analysis will be done with data sets ready made for teaching. The resulting
knowledge of statistics differs from university to university, but usually it
is sufficient for at least simple tasks in data analysis. Actually starting
with an analysis of one’s own, however, all too often is a bit rocky. The
reason for this being that just knowing statistics is not enough. Why is that
so? Self-collected data usually do not come as ready-made as it appears in
statistics classes. They need to be adapted, transformed, aggregated or
disaggregated, thoroughly documented and saved in meaning- and useful
partitions. This involves a large number of tiny steps and decisions in the
work process. It is all too easy to lose track of what has been done when,
how, why and with which result. Why has X been filtered, why has Y been
aggregated the way it has been aggregated, and where does the correction in Z
come from and how has it been justified? After a few days it is often not
clear anymore, why a variable does now look the way it does. And be it for the
reason that the seed number for some random number generator has either not
even been set or at least not been saved. It gets much more inconvenient
later on, when an article is ready for publication and the journal is asking
for documentation or even a replication data set. Or, when an interested
reader is sending an e-mail, asking politely for more detailed information on
data preparation and analysis. Because it is now that the search starts for
information that has been lost along the way.

Many of those
problems can be avoided by a well-conceived and thoroughly developed plan for
data manipulation and analysis accompanied by extensive documentation of
every step in the work process. Most if not all of the things one has been
doing can be kept within reach when the steps in the work process are clear
(and standardized to the extent possible), and when saving and documenting
becomes part of the daily routine of working with data.

This course will try
to introduce students to such a working-method with data and at the same time
do the first steps of data manipulation and analysis with STATA. The aim is
not, however, to actually teach statistics. It is assumed that the students
already have at least a basic knowledge of statistics with at least some descriptive
and inferential methods known to everybody. Basic knowledge of regression
analysis is mandatory.

The structure of STATA’s
command language, work process and documentation will be presented. The course
teaches some tricks how to achieve almost directly a publishable output. A
special focus will be put on graphical output and its improvement.

You may bring your
own data into the workshop. However, having data of one’s own is not
required.

Lernziel

You will get to
know:



- STATA’s user interface

- basic knowledge in data handling and data manipulation with STATA

- basic knowledge of the structure and workings of STATA’s commands for data
analysis

- basic knowledge of how to quickly produce publishable output with STATA

- basic knowledge of efficient process management and documentation using
STATA

After all, the
course is a language course of sorts. You will get to know STATA as a
language to code your data analysis.

Vorgehen

The course will be
held in a computer lab. All steps will be demonstrated by the instructor and
directly applied by the students. There will be room for free but guided
exercise.

There will be a
brief (90 minutes) intro into the Graphical User Interface of STATA for
people who have never worked with STATA before on the first day of the course
week. The course itself starts on the second day of that week, so that
newbies have an afternoon’s time to achieve a little acquaintance based on
the intro in the morning.

Literatur

As introductory
literature and a good guide book for further work I suggest Kohler/Kreuter
„Data analysis using STATA“ (meanwhile in its third edition) or K/K
„Datenanalyse mit STATA“ (the fourth edition from 2012 is strongly
recommended).

Hinweise zur Prüfung

In order to earn
credits students will have to complete a homework of about 5-15 hours
(depending on individual work pace and prior knowledge) with some data
manipulation and analysis, the production of some output that is (almost)
ready for publication and a thorough documentation of the whole process.