Homework 07

Repeated Measures ANOVA and Multilevel (MIXED) Linear Models

For Homework 07, you will be using the HELP dataset, learn more at:

• https://melindahiggins2000.github.io/N736Fall2017_HELPdataset/ &
• https://github.com/melindahiggins2000/N736Fall2017_HELPdataset

Refer to the repeated measures ANOVA and MLM lessons 21, 22 and 23. See

• Lesson 21 https://github.com/melindahiggins2000/N736Fall2017_lesson21
• Lesson 22 https://github.com/melindahiggins2000/N736Fall2017_lesson22
• Lesson 23 https://github.com/melindahiggins2000/N736Fall2017_lesson23

For the HELP dataset:

• OUTCOME VARIABLE: consider the 5 measurements of depressive symptoms (CESD) at baseline, 6m, 12m, 18m and 24m. Use variables cesd, cesd1, cesd2, cesd3, cesd4
• BETWEEN GROUP VARIABLE: Also consider the treatment group variable treat
• FYI: You will also need the id to
  1. properly restructure the data from WIDE to LONG format; and
  2. treat subjects as a random effect for the random intercepts approach to MLM

For Homework 7, complete the following:

1. Perform a repeated measures analysis of variance (RM-ANOVA) for the 5 CESD measurements across time by treatment group.
   1. treat time as a continuous variable (not as a factor) - this is your WITHIN group effect
   2. treat the treatment group treat as a factor - this is your BETWEEN group effect
   3. TABLE: present the table of the intercept, time, treat and time*treat interaction effects including the tests of significance. [Remember this significance might change depending on the treatment group coding - try flipping the 0 and 1 and run the model again to see if the significance changes]
   4. FIGURE: make a plot of the CESD means across time by group - if you can make it an error bar plot which has the means and CI’s (confidence intervals) or SE’s (standard errors)
2. Repeate the “repeated measures/longitudinal” analysis using a random intercepts MLM model
   1. REMEMBER to restructure the data from WIDE to LONG format
   2. TABLE: present the table of the intercept, time, treat and time*treat interaction effects including the tests of significance. [Remember this significance might change depending on the treatment group coding - try flipping the 0 and 1 and run the model again to see if the significance changes]
3. Compare the results between the 2 approaches
   1. compare the sample size differences
   2. why do you think the results are different or are similar?