Exploring Causality. Diet, Depression and BMI - Part 2 Question 2: Does diet affect depression? We will use the data from the same study we looked at in class. Here is the description of the study again: A group of researchers in Australia conducted a short (three-week) dietary study. In the study, 75 college-age students with elevated depression symptoms and relatively poor diet habits were randomly assigned to either a healthy diet intervention group or a control group. The researchers recorded the change over the three-week period on a numeric scale that measures depression (the CESD scale. The CESD (Centre for Epidemiological Studies Depression) score is based on clinical observations.) They also recorded body mass index (BMI) at the start and end of the 21-day period. The variables will be Group and Improvement. 1. Change columns in Starkey to explore this question. (Be sure to click the Quantitative first, then the Categorical.) Describe the parameter from the Starkey screen in plain language. (4 pc=] 3. Find and build your 95% CI. Show thought process. [5 pts]4. Interpret your Cl. (I am 95%% confident... etc) [3 pis] Does the evidence show that diet does make a difference in depression scores? Explain. [2 pts] Part 2 - Matched Pairs vs Randomized Comparative design. This experiment was set up using Marched Pairs design. That's because two depression scores were collected for each individual and we looked at the mean difference between those scores. (The variable Improvement was not directly measured, but rather arrived at through math, one column minus another.) Let's compare results with Randomized Comparative design, where we find Cls for and separately and see if they overlap. From bootstrapping a 1-variable Cl for Day 1 Score I get, or as my 95% confidence interval. 5. Find a 95% Cl for Day 21 Score. No need to show thought process. [3 pts]6. When compared to my interval, is this strong evidence that the scores improved? Explain. [4 pts] This is why Matched Pairs design is more powerful than Randomized Comparative experimental design, and should be used when possible. It is stronger to look at the mean offference mu(Day21- Day1), rather than the difference of means ( mu(Day21) - mu(Day1) ) [21 pts, total]