A perk of having a NERC funded studentship is priority attendance on the NERC Advanced training courses, and I was fortunate to gain a place on the Systematic review and meta-analysis for environmental sciences held at Royal Holloway University. Meta-analysis is a statistical technique used to combine results from different studies to identify patterns among studies, the strength of this is a higher statistical power is achieved than that of a single study. It was originally developed in medicine to gauge the effectiveness of treatments but is increasingly being applied to ecology.
We started off with lectures on the different types of reviews and an introduction to meta-analysis. The classic narrative review tells a story of the state of research in the area, the problem with this is there is no systematic search for literature, which can lead to bias as sources are only selected if that support only the view of the writer. A systematic review aims to take a qualitative approach, setting out the research question and standards for selecting literature before the review. A meta-analysis takes this further by producing a quantitative review by statistically combining and analysing the results of each study included in the review. A good meta-analysis starts with a systematic review were the search terms and inclusion criteria are explicitly stated, with the aim to be transparent, reproducible and avoid bias in publication as much as possible. In brief, a meta-analysis calculates an effect size for each study, which is the measure of strength for the phenomenon of interest, which are combined to give a overall average.
During the course we had a number of activities to using metafor package in R (a computer program commonly used by ecologists) and a mini-meta-analysis activity which was carried out in a group over the course of the week. Our project was looking at how disturbances in forests, including harvest and fire, affects soil microbial communities, and we were given papers related to this in advance of the course. The first step was to define the question we wanted to ask, using the PICO framework:
Having decided our question should be: How forest disturbances affect forest soil microbial communities as measured by microbial biomass and as compared with undisturbed forests. We then moved on to defining inclusion criteria for our mini-analysis and what metric we wanted to use to calculate the effect sizes. We decided to include papers which included comparisons of disturbed and undisturbed (i.e. the control) forests and a measurement of microbial biomass, which could be the amount of carbon or the amount of phospholipid fatty acids (PLFA). We then had to decide on which variables we wanted to investigate in our analysis and decided to look at the type of disturbance (fire, harvest or fertiliser) and the method of measuring microbial biomass. The next step was to extract the data from the papers into a spreadsheet which could be read into R and analysed using the metafor package, and then prepare a short presentation on our results.
The course was not all work and no play however, we were also given the opportunity to visit Royal Holloway’s picture gallery, which was established by the founder, Thomas Holloway, during the 1880s. I also spent some time exploring the grounds and admiring the founders building and there was even time to collect a few causal earthworm records!
This was the first NERC funded course I had attended (I am booked on two others this year) and I found it very worthwhile. It was not only a good introduction to the techniques of meta-analysis, aspects of which I will be using in my PhD research, but also helped improve my critical thinking. This was consolidated by re-reading a recent paper on a meta-analysis of earthworms and plant productivity (van Groenigen etl. al. 2014) which I then led a discussion on in the PREDICTS weekly journal club. After the course I was able to read the paper more critically and noted potential problems e.g. their inclusion criteria were not explicitly stated, as per meta-analysis ‘best practice’ than when I read it first time around.