Oswald Features
OSWALD: Object-oriented Software for the Analysis of Longitudinal
Data in S-plus
OSWALD is a suite of S-plus functions
and C/Fortran routines for reading, manipulating, displaying and
analysing longitudinal data. OSWALD has been in use in the
biostatistical and social science communities for more than two years,
and is currently the subject of a Economic and Social Research Council
grant (as part of the Analysis of Large and
Complex Datasets initiative) to extend its functionality in social
statistics applications. When completed in late 1997, OSWALD will
include the most of the following features:
Data Types and Data Manipulation
Oswald provides S-plus data types (classes) for longitudinal data,
including:
- Single time series with arbitrarily-spaced measurement points
- Unbalanced longitudinal data (collections of time series with
arbitrary measurement points)
- Balanced longitudinal data (collections of time series with common
measurement points)
- Longitudinal data with associated covariates (subject-specific,
time-specific, or general) [1]
- Event history data (such as life histories on individuals)
Oswald provides methods for reading, displaying and manipulating
longitudinal data, including:
- Reading from ASCII files
- Reading from file formats produced by other packages [2]
- Selecting and assigning subsets of longitudinal data (similar to
S-plus matrix operations)
- Selecting subsets of event-history data using a query language [2]
- Mathematical and summary operations on longitudinal data
- Sophisticated plotting (including `shadow' plotting for
informative graphs from large datasets and `trellis' plots)
- Interaction with longitudinal data plots (cf `identify')
- Visualisation of event-history data using the 3-D `lexis pencil'
model [1]
Data Analysis
Oswald is desgned to provide the user with a wide range of
techniques for the analysis of longitudinal data. These include:
Analysis of continuous longitudinal data
- Parametric model for longitudinal data (Diggle, 1988)
- Variogram estimation and display
- Informative dropout methods for longitudinal data (Diggle &
Kenward, 1994)
- Generalised random-effect models for longitudinal data (similar
to the functionality provided by SAS PROC MIXED) [1]
Analysis of discrete longitudinal data
- Generalised Estimating Equation (GEE) analysis (Liang & Zeger, 1986)
- With extensions to unbalanced/incomplete data sets [2]
- Alternating Logistic Regressions (ALR) for binary longitudinal data
(Carey, Zeger & Diggle, 1993)
- Generalised Linear Mixed Models [2]
- MLn analysis [2]
Analysis of binary and ordinal longitudinal data [2]
Many data sources in the social sciences provide event-history data
where many of the variables are binary or ordered categorical. Oswald
will be interfaced to the SABRE data
analysis system to provide analytic techniques for these types of
data, including:
- Logistic regression models for binary
- Models for count recurrent events
- Proportional odds models for ordinal events
- Mixed models (random effects), extending the above models
- Lagged response models
Availablility of Oswald
OSWALD is freely available on a trial basis under the terms of the GNU
General Public Licence for Windows and Unix systems. The distribution
includes the software (including full source code) and a 80-page
manual with fully worked examples. Users of Oswald are encouraged to
join the Oswald mailing list (send mail to
oswald-request@lists.lancs.ac.uk to join; there are presently
approximately 150 members) where assistance with using Oswald to
analyse data in practice may be sought.
Oswald is available on Unix and Windows platforms. S-plus is required
to be already installed to be able to use Oswald.
More information, and the software itself, may be obtained from the
OSWALD Web page at:
http://www.maths.lancs.ac.uk/Software/Oswald/
or by contacting the Oswald Development Team, Department of
Mathematics and Statistics, Lancaster University, Lancaster LA1 4YF,
UK. Oswald is also available on the Statlib archive of statistical software.
Footnotes
[1] Will appear in version 2.7
[2] Not present in version 2.7, but proposed to be
included in the final version
Dave M. Smith <D.M.Smith@lancaster.ac.uk>
Last modified: Thu Jun 5 12:12:26 BST 1997