MOST practice decision-making
Example 1: Pediatric obesity
Download Files
If you are interested in practicing fitting the ANOVA models, constructing interaction plots, and deciding which components to include in the hypothetical intervention, follow steps 1 –6 of the instructions below.
If you are interested only in practicing the decision-making process, skip steps (1) – (4), and begin on step 5.
Instructions
- Read the description of the experiment provided below.
- Download the artificial data set and import the data into your preferred statistical software package (e.g. SPSS, STATA).
- Estimate the ANOVA model, including main effects and interactions.
- Construct any interaction plots that may be useful in the decision-making process.
- Use the ANOVA output and relevant interaction plots to practice the decision-making process (you can use the file provided if you wish to skip steps 1—4).
- Check the answer key file to determine whether you chose one of the best combinations of components and component levels to build your hypothetical intervention.
Description of pediatric obesity example
A pediatric obesity researcher is interested in developing an intervention for families in which a child is obese. The researcher wishes to estimate the effects of five intervention components:
- VEG, aimed at increasing a child’s vegetable intake;
- ACT, aimed at increasing family physical activity levels;
- EDU, nutritional education for parents and children;
- DIN, aimed at increasing the frequency of homemade dinners;
- CHK, a home visit in which the nutritional habits of the household are reviewed and suggestions are made for healthier eating.
The researcher decides to use a 25 factorial design for the experiment, N=256. The first four factors correspond to the first four components. Each has 2 levels: yes (the family is provided with the component, coded 1 in the data set) and no (the family is not provided with the component, coded -1). The 5th factor corresponds to CHK, and has the following 2 levels: minimal (home visits occur once/month, coded -1 in the data set) and intensive (home visits occur biweekly, coded 1). The outcome variable is percent change from pretest to posttest in the child’s body mass index (BMI).