Meta-analysis: Past, present and future
Archived – Please note this event took place in the past and has been left here for reference purposes. View our full list of events to see what we have coming up or send us an email if there is a particular type of event you are interested in.
Date: 3 December
Venue: 6.300, University of Essex
Speaker: Elena Kulinskaya, School of Computing Sciences, University of East Anglia
Join Elena Kulinskaya for this seminar at University of Essex when she will talk about meta-analysis. Meta-analysis (MA) is a collection of statistical methods aimed at combination of evidence from independent studies. The origins of meta-analysis reach back at least to the beginning of the twentieth century, to work by Karl Pearson in 1904. Since the mid-seventies the term meta-analysis has become popular in several fields, among them medical statistics and the behavioural sciences. The most widely used procedures were perfected by mid-eighties.
The use of MA increased exponentially since
. MA deeply affected development of science and society, enabling the drive for evidence-based practice and policies. Unfortunately, in the eagerness to make meta-analytic methods accessible to end users, a kind of groupthink has taken hold of MA. This is slowly starting to change with appropriate statistical methods being developed
[2,3]. The future is exiting. Big Data, arising in many fields, are usually too big to be stored or analysed on one computer. Distributed computing stores and processes randomly subdivided datasets, subsequently merging the results
. However, routinely merging results is appropriate only for homogeneous data. Merging heterogeneous data, which differ systematically across datasets, can yield biased answers. Meta-analysts have developed methods for handling and synthesizing multiple datasets, investigating sources of heterogeneity, evaluating data quality, adjusting for possible bias, and reporting results. Coupling methods of Big Data analysis with MA techniques, appropriately adapted, will lead to important advances in Big Data applications.
1. Shadish, W.R. and Lecy, J. D. (2015) The meta-analytic big bang, Research Synthesis Methods,6, 246-264.
2. Kulinskaya, E., Morgenthaler, S., and Staudte, R. (2014) Combining statistical evidence, International Statistical Review, 82, 214–242.
3. Hoaglin D.C. (2015) Misunderstandings about Q and ‘Cochran’s Q test’ in meta-analysis, Statistics in Medicine, Early view, DOI: 10.1002/sim.6632
4. Chen, X. and Xie, M. A. split-and-conquer approach for analysis of extraordinarily large data, DIMACS Technical Report 2012-01, January 2012.