An experiment was carried out to assess the effect of Sweet Talk, a text messaging support system for patients with diabetes (“A Randomized Controlled Trial of Sweet Talk,” Diabetic Medicine [2006]: 1332–1338).

Participants in the experiment were 92 patients, age 8 to 18, with type I diabetes who had been on conventional insulin treatment for at least one year. Participants were assigned at random to one of three experimental groups:

Group 1: continued conventional insulin therapy Group 2: continued conventional insulin therapy with Sweet Talk support Group 3: followed a new intensive insulin therapy with Sweet Talk support One response variable was a measure of glucose concentration in the blood. There was no significant difference in glucose concentration between groups 1 and 2, but group 3 showed a significant improvement in this measure compared to groups 1 and 2.

a. Explain why it is not reasonable to attribute the observed improvement in group 3 compared to group 1 to the use of Sweet Talk, even though subjects were randomly assigned to the three experimental groups.

b. How would you modify this experiment so that you could tell if improvement in glucose concentration was attributable to the intensive insulin therapy, the use of Sweet Talk, or a combination of the two?

c. Draw a diagram showing the structure of the modified experiment from Part (b).