PART 1: CALCULATE AND INTERPRET MULTIPLE REGRESSION
1. Use the SPSS dataset Polit2SetC to answer the following research question:
What variables predict level of depression among low income women?
2. Use the following variables for this assignment:
CESD: depression score
AGE: chronological age
EDUCATN: educational attainment
INCOME: family income prior month
WORKNOW: current employment status
SF12PHYS: SF-12 physical health component score
SF12MENT: SF-12 mental health component score
3. Compute a multiple regression analysis, using the ‘exclude cases PAIRWISE option’.
4. Complete the table below based on the results of your regression analysis. Round to 2 decimals.
Table 1. Regression Model Predicting Depression in Low-Income Women
Predictor Variable b Standard Error Beta t p-value
Age
Education
Income
Employment status
Physical Health
Mental Health
R2 = ***
Adjusted R2 = ***
5. Write a brief paragraph summarizing the results using the provided template.
In this sample of low income women, (how many) of the (total number) predictor variables examined were significant predictors of depression level, after controlling for all other variables in the model (F = ?????, p= .???). Education (Beta = .06, p = .03), income (Beta =-.08, p = .01), work status (Beta = -.06, p = .03), XXX, and XXX (report remaining significant variables along with Beta and p-value) were significant predictors of women’s level of depression. (Higher/lower) education, (higher/lower) income, being (employed/unemployed), and having (better/worse) physical and mental health were associated with lower levels of depression. XXX (insert variable) was not a significant predictor of depression in this sample. Women’s score on (insert variable) was the strongest predictor of depression. Taken together, slightly (more/less) than half of the variance (adjusted R2 = .??, p < .??) in depression levels was explained by this set of predictor variables.
PART 2: CALCULATE AND INTERPRET LOGISTIC REGRESSION
1. Use Polit2SetB data set to answer the following research question:
Does smoking status predict health status, after controlling for age and BMI?
2. Compute a logistic regression model to examine if smoking (SMOKER) predicts the probability of being in good health (HEALTH) when other variables (AGE & BMI) are included in the model.
3. Use the following variables for this assignment:
Dependent variable HEALTH: self-rated health variable. This variable is coded as 0 = fair to poor health and 1 =good to excellent health.
Independent variables SMOKER: current smoker or not
BMI: respondent’s BMI
AGE: chronological age
QUESTION ANSWER
1. In the logistic regression analysis, how many cases were included in the analysis?
2. In the null model (Block 0), what percent of the cases were correctly classified?
3. In the null model (Block 0), how many cases (N) and what % of the included cases were predicted to be in good to excellent health?
4. In the full model (Block 1), what percentage of cases was correctly classified? Did the model classification improve when the 3 variables were included in the model?
5. In the full model (Block 1), how many cases were misclassified?
Describe the cases that were misclassified (i.e., they were predicted to be in one health group but they were observed to be in the other group). [HINT: read the frequencies in the body of the classification table like you would read a cross-tabulation table). Answer questions below based on the classification table.
6. How many cases were predicted to be in fair/poor health, but were observed to be in good/excellent health?
7. How many cases were predicted to be in good/excellent health but were observed to be in fair/poor health?
8. Was the most common misclassification (overestimation or under-estimation) of how many women would be in good/excellent health?
9. Of those who were observed to be in good to excellent health, what percent were correctly classified in the predictive model?
Interpret the results.
10. What effect does smoking have on women’s health status, after controlling for other variables in the model? Use the template below to summarize the results of the model.
The purpose of this analyses was to determine if smoking predicted health status, after controlling for age and BMI. The results of logistic regresssion indicated that the odds ratio for smoking in this analysis was (?) (95% CI: ?). This suggests that, with other variables controlled, the odds of being in good/excellent health declined by about (?%) for women who smoked. Based on the CI, there is a (?%) reduction in the odds of being in good/excellent health at worst and a (?%) reduction in the odds of being in good/excellent health at best for smokers. The 95% CI (does/does not) include 1.0, indicating (statistical significance/lack of statistical significance). Of the three predictor variables in the model, smoking had the (strongest/weakest) effect on the odds of being in good/excellent health. The other variables had a smaller effect on the odds: BMI (OR = ?) and AGE (OR = ?). Thus, every unit increase in BMI was associated with a (?%) reduction in the odds of being in good health. Every year of advancing age was associated with an (?%) reduction in the odds of being in good health.
