Here are some words from our Senior Entomologist, Dr Andrew Sailsbury, who has worked on the project from its outset...
With more than 80, 000 invertebrates recorded on the plots, this is not a simple task and is broken down into smaller processes. The first step is to organise the data.
The first group to be analysed is the nearly 8 000 pollinating insects recorded.
Data is entered into a spreadsheet, following each pollinator recording event (19 spreadsheets in total). We begin by combining this raw data into one spreadsheet and format it so that it can be read by statistical analysis software. This must be done with no transcription errors; not easy when dealing with spreadsheets containing more than 53, 000 cells. In addition the measure of flower number made on each recording event must be calculated and added to the spread sheet, to be used as an additional explanatory variable.
The next step is to decide on which groups of invertebrates it is appropriate to analyse - e.g. with pollinators we are looking at all the insects that were recorded visiting flowers - but also breaking this down into groups and species. This will allow us, for example, to find which plots bumblebees, honeybees and hoverflies tended to visit.
After many hours at a computer, the data is ready for analysis. Initially we will be carrying out analysis of variance (ANOVA) – an appropriate technique when dealing with a designed experiment of this type. However, with complex data sets the exact method of analysis may change. Analysis will only give a list of numbers and let us know what differences between the plots are significant. Consequently, this is only the first step in presenting the results which will need further interpretation.
It is also worth saying, this is just one dataset – similar data preparation needs to be carried out for the data from the other sampling methods used over the four years; pitfall traps, vortis suction sampler and gastropod traps.
This is when the RHS scientists will be in danger of going 'bug eyed'!