For this week’s practice assignment you’ll be working a simulated dataset dealing with the influences of three independent variables on the probability of quitting smoking. The dataset contains information on people who were smokers at baseline. The dependent variable is whether each person successfully quit smoking, and is assessed at twelve months after baseline. The three independent variables are (a) self-efficacy to quit smoking, (b) severity of addiction to nicotine, and (c) having (versus not having) a friend who quit smoking over the past 12 months. All three of these were assessed at baseline. Please download the dataset DataAssign10.sav and the corresponding codebook CodebookDataAssign10.doc from the Practice Assignment 10 page of the Assignments section of this course.
(1) Set up and run a logistic regression model with quitting smoking as the dependent variable and self-efficacy to quit as the sole independent variable.
(2) There is reason to believe that the bivariate association between self-efficacy to quit and actually quitting may overstate the effect of self-efficacy on the probability of quitting because of possible confounding. One potential confounder is the severity of an individual’s addiction to nicotine. Some smokers may be more severely addicted to nicotine than others. Those who are less severely addicted may sense that their addiction is not so severe, and this may translate into higher levels of self-efficacy. Conversely, those who are more severely addicted may sense that their addiction is severe, which may translate into lower levels of self-efficacy to quit. At the same time, nicotine addiction may directly affect individuals’ likelihood of quitting successfully. Those who are more addicted may have a harder time quitting than those who are less severely addicted. Thus, when we compare the quit rates of people with high self-efficacy to those of people with low self-efficacy, we may implicitly be comparing less severely addicted people to more severely addicted people. To address the potential confounding role of nicotine addiction in the association between self-efficacy to quit and the likelihood of quitting successfully, run a logistic regression model with quitting as the dependent variable, and with self-efficacy to quit and the nicotine addiction index as the independent variables.
(3) Another hypothesis is that having a friend who quit smoking makes individuals more likely to quit successfully. This could occur through a variety of mechanisms. Set up a logistic regression model in which quitting is the dependent variable and having a friend who quit in the last twelve months is the sole independent variable.
(4) Write a brief report that summarizes the results you obtained in Parts 1, 2, and 3. For Part 1, how do the odds of quitting smoking vary in relation to self-efficacy to quit? For Part 2, how are your conclusions about the effect of self-efficacy on the odds of quitting affected by adjusting for variations in the severity of nicotine addiction? To what extent is the association between selfefficacy to quit and actually quitting attributable to confounding by nicotine addiction? For Part 3, how do the odds of quitting smoking vary according to whether or not a smoker has a friend who quit?
In the Assignments section, please upload the following items to the Practice Assignment 10 page 24 hours before live session 10:
1. Your syntax file for carrying out the analyses requested in Parts 1, 2, and 3 above;
2. Your written report as described in Part 4.