Since speaking to you last, our 3000 pots have been joined by a further experiment containing 7100 pots. As you may remember from my last post Big experiments!! we are only really happiest when there are pots numbering into 4 figures!! But this many individuals in science causes a problem: the importance of being random.You quickly learn in science at school and then university and then in your job that you can’t just plonk a pot wherever you like in the glasshouse. At school you learn the basic principle that for scientific results to be ‘true’ you need replicates in your experiment and ideally should repeat it many times. Then at University you learn in a bit more depth of putting these replicates into a block that can be a tray or set of pots/containers and that a rep could be in a tray or a whole glasshouse. And then when you actually have to do it for real in your job you get statisticians involved and they start talking about “randomised block design” and “Latin squares” and they send you a crazy spreadsheet with everything “nested” in amongst each other.
Of course I totally understand that this is because we are trying to minimise the risk of affected results due to environmental conditions, for instance making sure that all the plants of one population are not next to the window. However, our experiments are so big that our randomisations become quite complicated. For the 7100 pot experiment, each pot is randomly assigned a position in a tray, then each tray is randomly assigned a place within a glasshouse and each glasshouse contains a rep.
Once the randomisation has been designed by the statisticians then it’s up to us to put it into action. And as you can imagine for the size of these experiments the task is quite a large one. For the experiment on the sandbeds it was easier to sow the plants in population order so that if we needed to replace a seedling we could do so easily. However to then put these pots into the random order, we effectively have to empty the beds and place them back on again, which is slightly soul destroying once you have done this 7 times but a job that Team Efficient has worked out can take two hours per rep. For the experiment in the glasshouse all the plants were again sown by population number and then we had a thick document listing all the pots telling us which glasshouse then which tray and which position in tray they went in. I worked out I could put out the 49 pots required for each population in 13 minutes and it took three of us three full days to sow that experiment.
Then last week we needed to spray the plants because they are part of our Quantitative Genetics experiment (I’ll explain our genetics stuff in a future post) and we wanted to test the populations (this is only a third of them, we will be running this experiment twice more) with different herbicides. The problem is that we have to take the pots out of the tray to spray them and then when we put them back in we haven’t got time or the memory to remember where they all go.
So I think you can guess what I will be doing for a couple of days next week – yep you’ve got it…….re-randomising!!!