You can devise a study using what you learned from the novice's temperature-aggression study described on page 39. Like the novice, you will want to design a study that tests a hypothesis.
Unlike the novice, you will probably not test the hypothesis that increasing temperature would increase aggression. You might draw your hypothesis from a question you have (e.g., "Does expressing gratitude make one happier? "), from a proverb (e.g., Is it true that "When you are right, no one remembers; when you are wrong, no one forgets?"), or from one of the many techniques described in Chapters 3 and 4.
For now, suppose you tested the hypothesis that expressing gratitude increases happiness. As you learned on page 48, you would want to do an experiment to test this hypothesis. If you had a two-group experiment, one group would be a "gratitude" group and the other would probably be a "no gratitude" group.
The key to making this experiment internally valid would be to randomly assign participants to either the gratitude or no gratitude condition. To learn how you could randomly assign participants, see pages 693-694 of your text or go here.
To help your study's construct validity, you would want your "gratitude" group to do something that would involve expressing gratitude (e.g., "each night before you go to bed, write down three things you are thankful for") and have the "no gratitude" group do something similar that did not involve expressing gratitude (e.g., "each night, before you go to bed, write down three things that happened to you today."). To improve construct validity, you would also want a good measure of happiness. The following link would get you to some measures of happiness.