Project: Correlated Data Analysis
Use the dataset m4project2016.dta for this project. This dataset is from a clinical trial of the use of estrogen patches in the treatment of postnatal depression. A total of 61 women with major depression beginning within 3-months of childbirth and persisting up to 18-months postnatal were randomly assigned to estrogen treatment (n=34) or placebo (n=27). The Edinburgh postnatal depression scale (EPDS) was administered at baseline (pretreatment) and monthly for 6-months after treatment. The EPDS has possible scores from 0 to 30 with a score of 10 or greater interpreted as possible depression.
The objective is to determine whether the estrogen patch is effective at reducing depression compared with the placebo (use p < 0.05 to decide significance).
The dataset m4project2016.dtahas the following variables:
• subj - patient identifier
• group - treatment group, 1=estrogen and 0=placebo
• pre - baseline EPDS score
• dep1 to dep6 - EPDS scores for months 1 to 6
Data Management:
1. Missing values for EPDS scores are coded as "-9" in the dataset. Recode these to "."
2. Use the user-written program misschk to evaluate the patterns of missing data for EPDS measurements. Type "finditmisschk" in the command window to get the program and then install. Then type "help misschk" in the command window to get a window with instructions on using misschk. Use the command shown to generate patterns of missing data:
misschkpre dep1-dep6.
Describe what the missing data patterns for dep1 to dep6. What is a benefit of random effects models compared to repeated measures ANOVA or MANOVA for analyses with missing data?
3. Duplicate the baseline EPDS score as a variable called dep0. Keep both the original baseline variable (pre) and the new dep0 variable.
4. The repeated measurements are in wide format (dep0 to dep6). Reshape to a long format with a single variable depfor the EPDS measurements and a new variablemonth to indicate the month (also keeping the other variables pre, subj, and group). This is to be done using the reshapecommand in Stata.Create variable labels for all variables, and value labels for categorical variables. The project do-file should be clearly annotated for this step.
Descriptive Analysis:
5. Evaluate the bivariate association between intervention status and baseline EPDS scores using a two-sample t-test. Report the mean and standard deviation for the EPDS score and p-value.
6. Make a graph of mean EPDS score by intervention status by month using the following Stata commands (feel free to edit the commands to customize the graph). Describe/interpret the graph.
anova dep i.month#i.group
marginsi.month#i.group
marginsplot ,xlabel(,labsize(small)) ytitle("EPDS Score") xtitle("Month") ///
title("Mean EPDS Score by Month and Group") name(linear, replace)
7. Make a graph of individual EPDS profiles by month (spaghetti plot) for 8 intervention and 8 placebo subjects (randomly chosen) using the following Stata commands (feel free to edit the commands to customize the graph). Describe/interpret the graph.
set seed 19493 /* change the seed number to select different observations */
generaterandno = runiform() if month == 0
bysort group (month randno): generate flag = _n <= 8 if month == 0
bysort subj (randno flag): replace flag=flag[1]
sort subj month
twoway connected dep month if flag == 1, connect(L) by(group) ytitle("EPDS Score") ///
xtitle("Month") msize(vsmall) xlabel(,labsize(small)) ylabel(,labsize(small))
Regression Analysis:
8. Use a linear mixed effect model (xtmixedor mixed) to evaluate the changes in EPDS with treatment and time. Use the following terms/options in the model:
a. Include intervention status (use i.group).
b. Use time post-treatment as a categorical variable (use i.month).
c. Include interaction of group and time post-treatment (use i.month#i.group).
d. Pretreatment EPDS score (use pre).
e. Random effects for subjects.
f. Fit the model using restricted maximum likelihood (REML).
9. Use the margins command to generate the means and standard errors for EPDS by intervention status and time post-treatment. Put the results in a Table suitable for publication and write a description of the analysis and results.
Attachment:- m4project2016v13.rar