Homework 5

Due Thurs 14 Oct in class

Assignment:

This assignment contains only one homework problem:

1. As usual, go thru all the conceptual exercises in Chapter 3.
2. Chapter 3 Exercise: 26
   
   A few Splus hints for Ex 26
   
   • Go to the data link on the course homepage and get file EX0326.ASC. Don't enter it by hand on the computer.
   • You'll need to make a couple of transformed response vectors. It's easy here's how to do it. Assume you've read the data into a dataframe called Ex0326.

      
     > Ex0326$logit_log(Ex0326$removed/(1 - Ex0326$removed))
     

     This new column Ex0326$logit holds the transformed values of the proportion removed. For reciprical you do:

     > Ex0326$recip_1/Ex0326$removed
     

     Since we're using t-procedures here assuming same variance in each group, it's important that the sd's for the two groups look similar. The disparity in sample size (35 and 12) makes it important that the sd's are fairly close. If you like, you can try running the t-test without assuming equal variances. This should be more reliable.
   • Remember for boxplot, you want to use code as the splitting variable. The 1s are the Honeybee Workers. You can also use the command line to get the boxplots:

     > boxplot(split(Ex0326$logit,Ex0326$code),
       ylab="logit of proportion of pollen removed",
       xlab="type of bee",names=c("honeybee workers","bumblebee queens"))
     

   • A rule of thumb regarding transformations: Transform only if you have to. Once you change scales, it gets very hard to interpret results.