Introduction to Meta-Analysis

But there are a variety of commercially available meta-analysis software packages that allow you to do it into those packages that give you answers more quickly and perhaps more easily. Now, what do we actually-- first thing we try to look at is the effect size for each study [effect size] in our big set of studies. Now, a regular effect size indicates a relation between two variables, but sometimes there. So if I ask people on a scale of 1 to how happy are you after owning a cat versus somebody on a scale of 1 to 50, 5 points higher means something different on each one.

So a standardized effect size allows us to standardize across all those different measurement metrics and compare the studies across each other. It allows us to average them later. And the other thing we might want to look at. There's a variety of them. The three most common in my experience are odds ratios, Pearson Correlations, and Cohen's d. Now, some articles report these for you.

Looks like you do not have access to this content.

And all you have to do is copy then right down. The article will report a correlation of 0. But sometimes they don't report those effect sizes. You have to actually hand calculate them. Now, any good meta-analysis software or meta-analysis book can offer you a variety of formulas to help you convert those.

For example, the t-test can be easily converted to correlation, as can a set of mean to standard deviations. Now, in addition to recording standardized effect sizes for each study, you also have to record the sample size so we can weight that later. But also you want to record any moderators that might change the size that effect, at least. For one example, one might try to say, OK. Well, how are we measuring how happy cat owners are or non-cat owners? We might just ask them, knock on door, scale of 1 to You might peer into their windows and see how happy do they look in there.

You might look at their social media and see how happy their posts are. And perhaps these different measurement methods might give us different kinds of answers.

For example, we might also look at the sample itself. Are they cat owners in the United States, cat owners from other countries? Are they people who have had cats for a long time? Is this their first cat? There are a variety of different ways to try to break up these types of studies to see whether or not these other variables might affect the relation between these two variables. I conducted meta-analysis once on whether or not strong arguments are more persuasive than weak arguments. And one of the key moderators was, well,. If they aren't thinking too carefully, the research tended to notice that they didn't actually respond to a strong argument better than weak argument.

One meta-analysis even found that people working at particular universities who researched it at that university found the thing they expected, but researchers at other universities didn't find the thing they expected,. So for each study, you might want to record any kind of information you think might moderate the effectiveness of that particular relationship between these two variables.

The overall effect, the weighted sample size average,. Usually research suggests that regardless of which meta-analysis method you use, you find about the same answer. So first thing is that sample size weighted average effect. Just like if you heard a poll that asked 50 people who they were going to vote for, you wouldn't trust it as much if you asked 10, people.

So because of that, we tend to weight the bigger sample sizes for more in that average. Now, the next thing is a little bit more complicated, but in that average you might also try to adjust for certain artifacts.

An Introduction to Meta-Analysis

So you can actually try to adjust those statistically. It's a little more complicated, and Hunter and Schmidt can walk you through that, but it's ultimately worth it to get a more accurate view of what's. The other main statistic one calculates is a homogeneity of variance. So basically, whenever you conduct a whole bunch of studies, we know through statistical theory they're going to vary. That is, some are going to find a stronger effect than others.

That's just going to be chance. If you had balls in a bag and 50 were green and 50 were red, and you only pulled out 10,. But sometimes you might get, say, seven green, three red or maybe even 10 green every great once in a while. So we know that's just going to happen by chance, and we can estimate how often that will happen.

So homogeneity test tells us is all this variation just chance or is it perhaps due to some sort of other factor, one of those moderators we talked about? Now these various homogeneity of variance statistics will actually give you an estimate of should we go on with those moderators or do we have an overall fixed effect, otherwise known as a homogeneous effect? So ultimately, we might see a homogeneous effect. It's a fixed effect that just says, OK, it's just random chance. We've got a pretty sure idea that this is the overall effect. We introduce the merits of meta-analysis and how it can form an important and informative part of a systematic review.

We explain the most common statistical methods for conducting a meta-analysis and common issues that may be encountered along the way. At the end of the day, delegates should be able to conduct a meta-analysis of their own and interpret the results of meta-analyses published in journal articles. Related topics that we don't cover on this course are 1 how to conduct a systematic search of the literature, and 2 assessing the quality of studies in a meta-analysis.

A basic level of statistical literacy is required as a prerequisite. In particular, delegates should have a basic understanding of standard errors, p-values and confidence intervals. On the 2nd, optional, half-day of the course, the theory of day 1 is put in practice with the use of R Rstudio and real-world datasets. A basic knowledge of R programming is recommended as a prerequisite taught on our 1 day course - 'Introduction to R'.

If you are not sure whether you have sufficient knowledge in R, please take our short test in the separate tab titled 'Prerequisite test for R workshop'. Please note, this category does not include hospital staff unless you hold an official contract with the university.

Finally, please note that no refunds will be given for non-attendance or cancellations made within 5 working days of the start of the course. This fee is needed to cover printing, catering, etc. She has authored several meta-analyses as well as articles on methodological issues in the area, and made numerous presentations on the topic.

Having contributed chapters to two books on meta-analysis, she co-edited Publication Bias in Meta-Analysis. Larry Hedges, University of Chicago A pioneer in meta-analysis, Professor Hedges has published over 80 papers in the area many describing techniques he himself developed, that are now used as standard , co-edited the Handbook for Synthesis Research , and co-authored three books on the topic including the seminal Statistical Methods for Meta-Analysis.

An Introduction to Meta-Analysis - SAGE Research Methods

He has also taught numerous short courses on meta-analysis sponsored by various international organizations such as the ASA. He works closely with the Cochrane Collaboration and is an editor of the Cochrane Handbook.

He has much experience of teaching meta-analysis, both at Cambridge University and, by invitation, around the world. Request permission to reuse content from this site. Added to Your Shopping Cart. Description This book provides a clear and thorough introduction to meta-analysis, the process of synthesizing data from a series of separate studies.

Meta-analysis has become a critically important tool in fields as diverse as medicine, pharmacology, epidemiology, education, psychology, business, and ecology. Permissions Request permission to reuse content from this site. Introduction Are the studies similar enough to combine?