Discuss the below:
Let's say you want to know if the amount of candy adults eat influences the number of times they brush their teeth in a day. Your independent variable is candy - you will manipulate how many pieces of candy they eat. Your dependent variable is the number of times they brush their teeth in a day. This is a ratio scale: they can brush their teeth zero times/day, and brushing their teeth 6 times/day is twice as often as 3 times/day. You could make as many groups as you'd like for this, with each group being supplied with and asked to eat a different amount of candy each day. You could select your subjects from a subject pool of undergraduates at a university Psychology department; make sure you randomly assign them to the groups below. Imagine you do the following:
Group 1: given no candy
Group 2: given 10 Smarties and 2 pieces of licorice
Group 3: given 50 Smarties and 10 pieces of licorice
Group 4: given 100 Smarties and 20 pieces of licorice
You would have to ask your subjects to not eat any other sources of candy, and to diary every time they brushed their teeth as well as what they ate every day, so that you could control for average sugar intake. However, with random assignment to groups, this should be quite balanced. Your hypotheses would be:
Null: all four groups will brush their teeth the same number of times/day
Alternative: all four groups will not brush their teeth the same number of times/day
Numerically:
Null: MU1=MU2=MU3=MU4
Alternative: all groups are not equal
Note that it would be poor form to make your alternative hypothesis directional; that would be more appropriate for a regression analysis.
For now, you only need two groups: 1. candy and 2. no candy
You probably need twenty or so in each group. You would make a one tail hypothesis.
What would it be?
You might want to record time brushing teeth rather than frequency, since the distribution of frequency will have a small range and may not be normal.
What do you think?