Q). You have developed a new type of flea treatment ointment for dogs and are setting up a placebo controlled randomised trial. Your plan is to do a head count of fleas on each dog, then apply either the real treatment or a placebo ointment, wait a few days, and then recount the fleas. Your response variable will be the decrease in fleas.

So far you have recruited 4 dogs to take part in a pilot study: two long-haired dogs and two short-haired dogs. You suspect that there might be a difference in the effectiveness of the treatment for long-haired versus short-haired dogs. You have two options in mind for an experimental design.

Option 1: Randomly pick two dogs for the treatment group and two dogs for the control group.

Option 2: Randomly pick one each of the long and short-haired dogs for the control group and one each of the long and short-haired dogs for the treatment group.

1. Why is it better to record the decrease in fleas based on a before and after count rather than simply count the number of fleas a few days after treatment?
2. Will either of these options allow you to test for an interaction effect between hair type (short or long) and treatment (real ointment or placebo)? Why or why not?
3. Which option do you prefer? Why?
4. What R command/s could you use to decide if the treatment has helped given the following two models have been fit?

fit1 <- lm(D_fleas ~ HairType)

fit2 <- lm(D_fleas ~ Treatment + HairType)

Questions are for like 2 marks each so just need a correct answer and a short explanation.