Statistical Meta-Analysis with Applications

Loading Map....


USQ Logo

SSAI and the University of Southern Queensland are pleased to announce the following workshop:

Statistical Meta-Analysis with Applications

16-17 June 2015, University of Southern Queensland, Ipswich Campus

presented by

Professor Bimal Sinha and Professor Suhail A Doi

About the presenters:

Professor Bimal Sinha is the University of Maryland Baltimore County (UMBC)’s Presidential Research Professor; a Fellow of the American Statistical Association; a Fellow of the Institute of Mathematical Statistics; author of many books and more than 150 scholarly journal articles; an editorial board member of many international journals. He has supervised/advised thirty (30) doctoral students.

Professor Suhail A Doi is from the Research School of Public Health, Australian National University, Australia, and a former academic of University of Queensland. As an Epidemiologist, he is a well-known authority on meta-analysis and a proposer of inverse variance heterogeneity (IVhet) and quality effect models. His presentation will cover these new methods in meta-analysis.

About the workshop:

Statistical meta-analysis deals with a variety of sophisticated statistical methods to efficiently combine the results of several studies all with a common target. Examples of such studies abound in the literature. Some common application areas include gender studies in education, EPA studies of effects of second hand smoking on women, and controlled or comparative trials in medicine and epidemiology. In this course, we will describe the basic concepts of effect size for continuous measurements as well as qualitative attributes, combination of tests and estimates of effect size, tests for homogeneity of effect sizes, analysis of one-way random effects models, meta-analysis of binary data, meta-regression and publication bias. The common data situation in meta-analysis is the availability of only published data like effect size estimate plus standard error or effect size estimate plus confidence interval. Many real data sets of this type will be presented and analysed covering the fields of educational research and health sciences. Some computational aspects of meta-analysis using standard statistical software like R and SAS will also be mentioned.

Meta-analysis, a term coined by Glass (1976), is intended to provide the statistical analysis of a large collection of analysis results from individual studies for the purpose of integrating the findings. Meta-analysis, or Research Synthesis, or Research Integration is precisely a scientific method to accomplish this goal by applying sound statistical procedures, and indeed it has a long and old history. Although some early works of Pearson (1904), Tippett 1931), Fisher (1932), Cochran (1937), Cochran and Yates (1938), and Mosteller and Bush (1954) provided a logical foundation for meta analysis, the appearance of several books, notably Glass et al. (1981), Hunter et al. (1982), Rosenthal (1984), Hedges and Olkin (1985), and the edited volume by Cooper and Hedges (1994) and literally thousands of meta-analytic papers during the last twenty years or so, primarily covering applications in health sciences and education, has made the subject to have a very special role in diverse fields of applications. The present short course will be based on the book: Statistical Meta-Analysis with Applications, John Wiley, 2008.

The notion of effect size is central to many meta analysis studies which often deal with comparing two treatments, control and experimental, in an effort to find out if there is a significant difference between the two. In the case of continuous measurements, a standardised mean difference plays an important role to measure such a difference. In the case of qualitative attributes, difference or ratio of two proportions, odds ratio and phi-coefficient are used to capture such differences. Of course, if the objective is to study relationship between two variables, an obvious choice is the usual correlation coefficient.

A fundamental assumption behind conducting a meta analysis or pooling of evidence across studies in order to obtain an average effect across all studies is that the size of the effect (basic parameter of interest) reported in each study is an estimate of a common effect size of the whole population of studies. It is therefore essential to test for homogeneity of population effect sizes across studies before conducting meta-analysis if obtaining an estimate of average effect or its test is the primary goal of meta-analysis.

Recent meta-analytic work however concentrates more on discovering and explaining variations in effect sizes rather than assuming that they remain the same across studies, which is perhaps rarely the case owing to uncontrollable differences in study contexts, designs, treatments, and subjects. This is precisely the spirit of some recent research in meta-analysis using random and mixed effects models, allowing inclusion of trial-specific covariates which may explain a part of the observed heterogeneity.

Given a very broad spectrum of topics that can be covered under the umbrella of a course on meta-analysis, our goal in this course is primarily concerned with some basic statistical aspects of meta-analysis.

Target Audience:

This workshop is aimed at statistics faculties, biostatisticians, epidemiologists, statistics graduate students and professional statisticians. A basic mathematical statistics background at the master’s level is a prerequisite to get the most out of this workshop.

Learning Objectives: It is expected that successful attendees will come out with a sound knowledge of basics of statistical meta-analysis, and will be able to carry out applications of meta-analysis with a variety of data sets.

Course Outline:

Topic 1. After a brief introduction to the subject of meta-analysis, a description of various standard measures of effect size based on means, proportions, phi-coefficient, odds ratio, and correlations will be provided. We will illustrate how to estimate these measures along with the properties of these estimates. We will then explain a) methods of combining individual tests of effect sizes based on combination of P-values, and b) methods of combining individual estimates of effect sizes. Some illustrative examples will be given.

Topic 2. In this lecture, we will discuss various tests of the important hypothesis of homogeneity of population effect sizes.

Topic 3. A special problem of drawing inference about a common mean of independent normal populations will be discussed.

Topic 4. This lecture is designed to address the methods of meta-analysis in case of sparse binary data. Some applications will be presented.

Topic 5. Explaining heterogeneity using a random effects model and meta-regression techniques will also be addressed.

Topic 6. In this concluding lecture, we will first consider methods for detecting publication bias. Finally, some computational aspects of statistical meta-analysis using R and SAS will be mentioned.

Topic 7:  Suhail Doi’s presentation on `Recent Advances in Meta-analysis’ lecture will cover the heterogeneity issue, and provide alternatives to random effects model that perform better in terms of re-distribution of weights and maintain the nominal level of confidence for the confidence interval of the common effect estimate.


Registration Fees:

Early Bird Fees (payment before 8 May 2015):

Full Fee: $700.00

SSAI Members:  $600.00

Students: $400.00

Registrations from 9 May 2015:

Full Fee: $800.00

SSAI Members:  $700.00

Students: $500.00


Travel Expenses

Occasionally workshops have to be cancelled due to a lack of subscription. Early registration ensures that this will not happen. Please contact the SSAI Office before making any travel arrangements to confirm that the workshop will go ahead, because the SSAI will not be held responsible for any travel or accommodation expenses incurred due to a workshop cancellation.

Cancellation Policy

Cancellations received prior to Friday, 15 May 2015 will be refunded in full. Confirmation of the refund having been processed will be emailed. Should additional documentation pertaining to the refund be required, a $20 administration fee will be charged.

After 15 May 2015 no part of the registration fee will be refunded. However, registrations are transferable within the same organisation. Please advise any changes to [email protected].






Statistical Meta-Analysis with Applications
When: 16/06/2015 - 17/06/2015
Time: 9:00 am - 4:00 pm
Cost: from $400.00
Location: University of Southern Queensland,
11 Salisbury Road,

Get the latest posts delivered to your mailbox:

Show Buttons
Hide Buttons