The slimming pill experiment is badly designed because if it is assumed that the mouse is the experimental unit, comparisons between groups could be biased. However, if the cage is the experimental unit, then there are not enough cages. So far we have not discussed how large an experiment needs to be. This is discussed in the section "improving the precision of your experiment". For the moment, just use your best guess of the most appropriate numbers.
Can you redesign it to overcome these limitations? First chose how many cages you would like in each treatment group and decide how many mice should be in each cage then press next.
You have chosen a structured animal house shelving configuration.
This is unfortunately not a good choice - go back and try again.
In this configuration all group 1 cages are stored on the top shelf and all group 2 cages are stored on the bottom shelf, or visa-versa. Problems exist with this configuration because top and bottom shelves may be different environment which could lead to bias.
You have chosen a systematic animal house shelving configuration.
This is unfortunately not a good choice - go back and try again.
In this configuration all groups 1 and 2 are stored on the same shelf in a systematic order 1, 2, 1, 2, etc. This could lead to bias if there is a temperature, light or ventilation gradient across the room.
You have chosen a randomised blocked animal house shelving configuration.
Well done!
In this configuration equal numbers of group 1 and group 2 are stored on each shelf, but in random order. This is a randomised block design (each shelf being a block) that is unbiased and efficient, even if the environment differs between shelves. However, you must have the correct statistical analysis for this design in order to remove the block differences. Randomised block designs are discussed in more detail in the 'Precision' section.
You have chosen a random animal house shelving configuration on 1 shelf.
Good choice - however, there is a better one - go back and try again.
In this configuration there is a random mix of group 1 and 2 cages on a single shelf. This avoids bias and is a good design if the shelf is large enough.
You have chosen a random animal house shelving configuration on 2 shelves.
Good choice - however, there is a better one - go back and try again.
In this configuration there is a random mix of group 1 and 2 cages on 2 shelves. Although this experiment would be unbiased, it may lack precision if environmental differences between shelves are large because this will increase the error due to differences between cages on the same treatment.
