The first day that I went out for data collection I noticed that there were sprinklers in operation at the garden. This was noticeably impacting the amount of bees I active in the garden compared to the last time I was there for my initial observations. Therefore, I decided that it would be best to come back another day to get a more accurate representation, preferably after it had rained because it’s less likely they would be using sprinklers to irrigate the garden then.
After running a couple days later, I was successfully able to use my haphazard sampling strategy (using a random number generator to determine paces then choosing the closest flowers of the corresponding colour to sample). This time I sampled all flowers in a quarter meter square quadrant as the experimental unit. What I saw after this first sampling date was in line with my hypothesis that yellow and purple flowers would be more attractive to bees than whites and pinks so that wasn’t particularly surprising. However, I was surprised by how consistently true this was over all my replicates after this first sampling date.
Going forward, I am going to change my experimental unit to a single flower of each colour as opposed to a quarter meter quadrant of flowers. I made this decision based on your suggestion that using a larger sample area could introduce sample bias if there are other colours represented in the area. A single flower is also more simple to observe and measure than an area of flowers. I think this will make my research more objective and results more accurate than if I were to continue using a quadrant as my experimental unit.