Docstoc

THE INVESTIGATOR

Document Sample
THE INVESTIGATOR Powered By Docstoc
					THE INVESTIGATOR
Training Program for Educational Research Workers


           Advanced Multivariate and Multilevel Analysis
                 Flinders University – EDES9005 and EDES9005A


Participants    Students working for PhD, EdD, MA and MEd degrees. Other
                students and interested persons are welcome

Staff           Professor John P. Keeves, Dr Kelvin Gregory, Mr Pawel Skuza
                and Dr I Gusti Ngurah Darmawan

Duration        One Semester

Class Contact   One 90 minute lecture and one 90 minute computer practical
                per week

Pre-requisite   An Introductory Course in Statistics

Units           6 units for MEd, 9 units for EdD

Location        Investigator Room, Room N105, Sturt Buildings, Flinders
                University

Time            Tuesdays afternoons, 1:30 pm to 4:30 pm, during Semester 2.
                First lesson starting on Tuesday 25 July 2006.

                This topic aims to prepare students and research workers in education and
Aim
                the social and behavioural sciences to select and employ appropriate
                analytical procedures for the examination of data collected in surveys, quasi-
                experimental research studies and longitudinal studies as well as to draw
                appropriate conclusions and interpret the research findings from such
                studies. The course concentrates on an understanding of and on the use of the
                analytical procedures of linear regression, path analysis, multiple regression,
                factor analysis, cluster analysis, canonical analysis, discriminant analysis,
                analysis of variance and covariance, latent variable path analysis, and
                structural equation modelling. In addition, the problems of multilevel
                analysis are examined and an understanding and experience in the use of the
                analytical procedure of hierarchical linear modelling is provided both for
                studies of growth and of school and classroom effects. The HLM, MLWin,
                MPlus and STREAMS programs are introduced as appropriate procedures
                for multilevel analysis. The implications of the choice of a particular
                multivariate analytical procedure for the design of quantitative research
                studies in the social and behavioural sciences are considered.
Course Outline
Class 1       Introduction to multivariate and multilevel analysis, Missing data
25 Jul 2006   Correlation coefficients and their meaning, Effects of aggregation and
              disaggregation of data, Handling missing data through imputation


              Exercise: Problem
              Influence on correlation matrices of
                 (a)    imputation of missing data
                 (b)    aggregation of data
                 (c)    disaggregation of data


Class 2       Least squares estimation, linear and multiple regression
1 Aug 2006    Introduction to path analyses, causal modelling
              The development, testing and interpretation of casual models


              Exercise: Problem
              Examination of direct and indirect effects in casual path models of
              different levels of aggregation
Class 3       Factor analysis, principal components, varimax rotation, orthogonal and
8 Aug 2006    oblique axes


              Exercise: Problem
              Examination of factor structure under different conditions using factor
              analysis
Class 4       Cluster analysis and multidimensional scaling
15 Aug 2006

              Exercise: Problem
              Examination of data to form clusters of variables and cases
Class 5       Canonical analysis, latent variables
22 Aug 2006

              Exercise: Problem
              Examination of canonical structure between two sets of variables
Class 6       Discriminant analysis for dichotomous and polychotomous variables
29 Aug 2006   as the criterion

              Exercise: Problem
              Examination of factors influencing a criterion variable with
              dichotomous and polychotomous categories
Class 7       Multivariate analysis of variance and covariance, repeated measures,
5 Sep 2006    relative change and absolute change over time
              Estimation of design effects

              Exercise: Problem
              Examination of the use of multivariate analysis of variance for
                 (a)    a covariance adjustments,
                 (b)    relative change,
                 (c)    absolute change,
                 (d)    to estimate design effects
Class 8       Latent variables, Partial least squares path analysis
12 Sep 2006   Jackknife estimates of sampling error


              Exercise: Problem
              Examination and testing of a latent variable path model with a simple
              random sample or a complex sample
Class 9       Maximum likelihood estimation
3 Oct 2006    Structural equation modelling using maximum likelihood estimation
              procedures using LISREL or AMOS


              Exercise: Problem
              Comparison of estimation using partial least squares analysis and
              maximum likelihood procedures
Class 10      Confirmatory factor analysis of test-item data using LISREL or
10 Oct 2006   AMOS
              Hierarchical and nested models


              Exercise: Problem
              Examination of factor structure of a mathematics test
Class 11      Analysis of growth and learning using hierarchical linear modelling
17 Oct 2006
              Exercise: Problem
              Estimation of growth of rats
              Estimation of learning in a classroom experimental study
Class 12      Hierarchical linear modelling and multilevel analysis
24 Oct 2006   Cross-level interaction effects, Value added effects, residuals


              Exercise: Problem
              The class-size problem and the effects of streaming using HLM
Class 13       HLM and the v-known estimation procedure
31 Oct 2006

               Exercise: Problem
               Meta analysis of gender effects in reading
               Measurement problems with gender effects in reading over time
Class 14       Multilevel Path Modelling, Suppressor variables, and cross-level
7 Nov 2006     interaction effects
               Use of STREAMS, AMOS and MPLUS


               Exercise: Problem
               Examination of a simple two level path model using MPLUS
Assignment

MEd: EDES9005

The assignment to be completed at the conclusion of the topic involves the demonstration
of competence in the use of two of the analytical procedures treated during the topic and
in the inferences drawn and the reporting of results of analyses undertaken through the
use of these procedures.

EdD: EDES9005A

The assignment to be completed at the conclusion of the topic involves demonstration of
competence in the use of three analytical procedures treated during the topic and on the
inferences drawn and the reporting of analyses undertaken through the use of these
procedures. Students are encouraged to analyse their own data.

In addition, all members of the class are expected to obtain working experience with all
analytical techniques considered in the topic.

Reference Book

Keeves, J.P. (ed) (1997) Educational Research, Methodology and Measurement. An
    International Handbook (2nd edn), Oxford: Pergamon.
.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:8
posted:9/6/2011
language:English
pages:4