Some breeds of dogs, like bloodhounds, have been bred to perform well on scent detection, and some breeds have not. So how do these different types of dogs compare on a scent detection test, and how do they compare to wolves? Researchers tried to answer these questions by comparing four groups of animals: dogs bred for scent detection, dogs bred for other purposes, dogs with exaggerated shortnose features like a bulldog, and wolves (Polgár et al., 2016). They tested 12 wolves that were raised in captivity and 41 dogs from the three groups. To test the animals, the researchers placed four ceramic flowerpots flipped over so there was a small hole at the top. Under each pot was a plastic container. Three of these containers were empty and one contained raw turkey meat. The animals were test 20 times each under various levels of difficulty. The easiest level had no lid on the plastic container. They then had harder levels with five, three, and one small hole on the lid of the plastic container. The most difficult level had no holes on the lid of the plastic container. Each dog was tested four times under each condition for a total of 20 times. The results of how many times each animal picked the correct container are in the file DogSmells.

a. Put the data into the Multiple Means applet. Which type of animal had the highest mean score? Which had the lowest? 

b. What is the value of the Mean Group Diff statistic? What is the value of the F-statistic? What is the simulation-based p-value for each?

c. There are two outliers in the dataset. One of the short-nose dogs and one of the dogs not bred for scent only got three correct. Perhaps these dogs were confused as to what to do. Remove them from the dataset. What is the new Mean Group Diff statistic and F-statistic? What is the simulation-based p-value for each? 

d. You should have found that the Mean Group Diff statistic decreased, while the F-statistic increased. Why did this happen?