Practical Data Management and Analysis for Public Health Assignment 6
Graded Assignment 6
Last week, we began looking at the pretest and posttest data in an effort to answer the question: “Did the intervention work?” We focused on four variables, all of them continuous: knowledge, self-efficacy, and motivation scale scores, and, most importantly, actual LDL levels.
Now we will approach the same question using two categorical dependent variables. The first is the dichotomous version of LDL, hildlb, which indicates for each participant whether she or he had high LDL levels (>160 mg/dL) following the intervention. Participants with high LDL received a 1 on this variable, and all others received a 0. The second is a four-level ordinal version of LDL, called ldl5catb. This variable would divide participants into the following five ordinal categories of LDL: optimal (coded 1), near optimal (coded 2), borderline high (coded 3), high (coded 4), and very high (coded 5). But no one in the sample had LDL levels enough to be categorized as optimal, so in fact ldl5catb takes four values: 2, 3, 4 and 5. Here we will essentially replicate the analyses we did last week, but this time focusing on these dichotomous and ordinal variables (and, for some analyses, their pretest counterparts: hildla and ldl5cata).
Please download the GunkPrePost2.sav. dataset from the Practice Assignment 6 page of the Assignments section of this course.
Assignment
(1) To begin, pretend that no baseline (pre-intervention) data were collected. In other words, pretend that we have a posttest-only control group design. Conduct an appropriate statistical test of the null hypothesis of no intervention effect on post-intervention LDL status: high versus not high. The analysis should produce three hypothesis tests.
(2) Pretending once again that no baseline (pre-intervention) data were collected, conduct an appropriate statistical test of the null hypothesis of no intervention effect on post-intervention LDL status: optimal, near optimal, borderline high, high, and very high. How does the statistical conclusion compare to those obtained using the dichotomous LDL variable?
(3) Next, pretend that there is no control group. In other words, limit your analysis to participants who were assigned to the treatment group (i.e., treat=1), and analyze the data as if they were obtained via a one-group pretest-posttest design. Conduct an appropriate statistical test of the null hypothesis of no intervention effect on post-intervention LDL status: high versus not high.
(4) Pretending once again that there was no control group, conduct an appropriate statistical test of the null hypothesis of no intervention effect on post-intervention LDL status: optimal, near optimal, borderline high, high, and very high. How does the statistical conclusion compare to that obtained using the dichotomous LDL variable?
(5) Collect all of the results form (1) through (4) above into a written report with two sections. In the first section of the report, cover analyses (1) and (2); in the second section cover analyses (3) and (4). For each analysis, please report (a) how big the effect was in quantitative terms (e.g., what was the relative risk of high LDL among intervention participants compared to control group members?), and (b) whether or not the effect was statistically significant at the .05 level. Use your discretion as to whether or not to use tables to facilitate your presentation of the results.
Submission
In the Assignments section, please upload the following items to the Practice Assignment 6 page 24 hours before live session 6:
1. Your syntax file for carrying out the analyses requested in items (1) through (5) above;
2. Your written report as described in (6).
