When studying pets in a different
context than their own home
(e.g., in a veterinary surgery) it is important to control for any
potential
sources of variation, so make sure that you plan carefully what time of
day/day of the week
you will be sampling.
When sampling from pets in different contexts, first, think about what your unit of replication is. This is not a trivial as it sounds (e.g., one vet treating 3 animals does not necessarily consitute n = 3!). A common mistake is to have all members of treatment group 1 in one context, and all members of treatment group 2 in another. The unit of replication in this case would be context (e.g., vet surgery) not individual (i.e., animals from one vet would constitute n = 1!). The reason this is important is because the variability within each context will usually be lower than the variability between each context (things like effects of the town in which (for example) the veterinary surgery is located, or even the vet surgeon themselves, etc. will come into play). So, any effect that you observe might be an effect of context, rather than an effect of your treatment.
One way to deal with this is to randomize your treatment allocation.
Animals in each treatment group should be age/sex matched as far as is possible.
Animals in each group should come from multiple owners, breeds (if appropriate) and be randomly allocated to treatment groups.
Try to sample from as many contexts as you can (e.g., three different veterinary surgeries in each of three different towns using three different vets in each surgery if possible!)
If you have animals from 2 (or fewer) surgeries, you could treat 'vet' as a variable and use it in the analysis to see if there are any systematic differences evident. If you are sampling from >2 surgeries, it may be possible to treat it as a random effect - we will deal with this in the data analysis sections.
When sampling from pets in different contexts, first, think about what your unit of replication is. This is not a trivial as it sounds (e.g., one vet treating 3 animals does not necessarily consitute n = 3!). A common mistake is to have all members of treatment group 1 in one context, and all members of treatment group 2 in another. The unit of replication in this case would be context (e.g., vet surgery) not individual (i.e., animals from one vet would constitute n = 1!). The reason this is important is because the variability within each context will usually be lower than the variability between each context (things like effects of the town in which (for example) the veterinary surgery is located, or even the vet surgeon themselves, etc. will come into play). So, any effect that you observe might be an effect of context, rather than an effect of your treatment.
One way to deal with this is to randomize your treatment allocation.
Animals in each treatment group should be age/sex matched as far as is possible.
Animals in each group should come from multiple owners, breeds (if appropriate) and be randomly allocated to treatment groups.
Try to sample from as many contexts as you can (e.g., three different veterinary surgeries in each of three different towns using three different vets in each surgery if possible!)
If you have animals from 2 (or fewer) surgeries, you could treat 'vet' as a variable and use it in the analysis to see if there are any systematic differences evident. If you are sampling from >2 surgeries, it may be possible to treat it as a random effect - we will deal with this in the data analysis sections.