Experimental Design – part 4 – Assigning groups


Don’t like the narrated version? Scroll down for slideshare slides

or, read the text version of this content here

Reach part 5 via this link

You are conducting an investigation in bacterial pathogens in the tick population of the UK red squirrel population. Ticks were collected from red squirrels from 9 sites across Scotland and England, 5 squirrels per location, 3 ticks per squirrel as part of another study. How would you plan to organise the data for analysis?
Summarise the data from the 3 ticks, then 5 squirrels from each site to generate a single value from each of the 9 sites, which you will then use in your final analysis. Sample size = 9
Yes, most likely this would be the best plan. Pretty clearly, the 3 ticks living on the same squirrel are very likely to be very similar to one another. However, the 5 squirrels also aren’t independent. They live in the same forest, interact with the same plants and other animals, same undergrowth, etc. Its very likely that the differences between squirrels from the same location are much smaller than the differences between different sites. When it comes to your final analysis you will be able to test how related these samples are.
Consider the data from each tick as your experimental units. Total sample size = 9 x 5 x 3 = 135
No, the three ticks live on the same squirrel, they share an environment, they’re not likely to be independent. This doesn’t mean you shouldn’t analyse them, just that you will combine the measurements from them
Consider the 5 squirrels as your experimental units, the 3 ticks as technical repeats. Total sample size is 9 x 5 = 45
Maybe, but probably not! Your data could prove that the 5 squirrels are independent of each other but I would assume that as the squirrels that live in the same forest are exposed to the same environment. When planning the experiment I would assume that I would pool the data from the 5 squirrels to generate a single value.
You are investigating the impact of reading prescription on visual field sensitivity. The clinician running the study wants you to test the study in the following way. Phase 1; perform perimetry with the subject is wearing their own reading glasses, phase 2 perform perimetry with the subject wearing perimeter-calculated reading addition, phase 3 perform perimetry with no addition. All three phases are to be conducted during the same testing session with a break of 5 mins between each test. Sample size of 50 subjects. Can you identify any potential problems with this set up? Think first, then click the answer
Potential criticism
There is a danger here of an order effect. Is the gap between tests sufficient that the second and third tests are not affected by the preceding tests? This may not be an issue at all, however, it would be very easy to randomise the order the participants took the tests so why not just remove the questions

reach part 5 via this link