Question

Instructions

• Please answer all questions
• Use the given dataset “Homework_EX_1”.
• Please include the relevant SPSS output (such as tables and graphs).
• Include the screen snapshots of how you specify the model in SPSS (use the “Print_screen” function in your keyboard to capture each screen and paste it in the word document).

Please provide a clear WELL-WRITTEN interpretation for each question.

Name	Variable information
Diabetes	Diabetes status (Yes, No)
Age	Age measured in years
Race	Race of each woman described as White, African American or Others.
Glucose	Blood glucose level measured in mmol/L
BMI	Body mass index measured in Kg/m2
Statins	Intake of Statins medication to lower the levels of LDL (Yes, No)
Alcohol	Intake of alcohol (Yes , No)
Smoking	Smoking cigarettes (Yes, No)
Exercise	Any type of exercise (i.e. physical activity) routine (Yes, No)
Physical_Activity	Level of physical activity (Minimum, Moderate and vigorous)

Objective 1: To assess relationship of body mass index (BMI) and relevant variables using dataset (Homework_EX_1). Data were collected on 2763 female to assess their risk of metabolic conditions. The variable Age, smoking, race, exercise, physical activity, alcohol, glucose, statins and diabetes are included in the given dataset.

Q1. First, determine the relationship between body mass index, BMI (dependent variable) and Blood glucose, glucose only (independent variable).

1. What is the null and alternative hypotheses?
2. Comment on the fit of the model? How much variability does it explains?
3. Is there a significant relationship between these two variables?        
4. Interpret the blood glucose coefficient?
5. Are the assumptions of the linear regression model satisfied?

Q2. First, determine the relationship between body mass index, BMI (dependent variable) and diabetes only (independent variable).

1. Is there a significant relationship between these two variables?
2. Interpret the diabetes coefficient?

Q3. Next, determine the relationship between body mass index, BMI (dependent variable) and glucose (independent variable) while including Age, diabetes, smoking, alcohol or physical activity variables in the model.

1. Fit regression model for body mass index using enter method for all these variables. If Age, diabetes, smoking, alcohol or physical activity are significant variables to keep in the model. If yes, then please specify the nature of this relationship i.e. confounder or effect modifier and also include interaction term, if appropriate.
2. Now, fit regression model for body mass index using different selection methods (stepwise, backward, forward) and compare their results?
3. Comment on the fit of the final chosen model using different selection methods? How much variability does it explains?
4. Is there significant relationship between these independent and the dependent variable? Which variables are significant and interpret these coefficients?
5. Are the assumptions of the linear regression model satisfied for your final chosen model?

Q4. Now, Compare the fit of the first model in Q1 and the final model in Q3? Does the inclusion of these variables improve the model? Should all of them be included in the model? Explain your reasoning?

Objective 2: To assess relationship of diabetes and relevant variables using dataset (Homework_EX_1). Data were collected on 2763 female to assess their risk of metabolic conditions. The variable Age, smoking, race, exercise, physical activity, alcohol, glucose, statins and body mass index are included in the given dataset.

Q5. To assess relationship between diabetes as dependent variable and other variables as independent variable.

1. Run a regression model to test for association between diabetes and blood glucose. Does this suggest that blood glucose is associated with diabetes? How do you interpret the effect size and 95% CI? Now, also include race variable in the model, does this suggest that race is associated with diabetes? How do you interpret the effect size and 95% CI of variable race when adjusted for blood glucose level?

2. Run a regression model to test for association between diabetes and independent variables (body mass index and blood glucose) and also their interaction term (body mass index and blood glucose). Is the interaction term significant? How do you interpret the interaction co-efficient effect sizes and their 95% CI?

3. Run a regression model to test for association between diabetes and statins. Does this suggest that statins are associated with diabetes? How do you interpret the effect size and 95% CI?

4. Run a regression model to test for association between diabetes and exercise. Does this suggest that blood glucose is associated with diabetes? How do you interpret the effect size and 95% CI?

5. Run a regression model to test for association between diabetes and Smoking. Does this suggest that smoking is associated with diabetes? How do you interpret the effect size and 95% CI?

6. Run a regression model to test for association between diabetes and alcohol. Does this suggest that alcohol is associated with having diabetes? How do you interpret the effect size and 95% CI?

Q6. Use forward LR method and enter method to assess relationship between diabetes as a dependent variable and all other co-variates (Age, blood glucose, race, smoking, alcohol, exercise, physical activity, and body mass index). Also include in the models to assess the following three mentioned interaction terms; body mass index and blood glucose, alcohol and smoking and finally body mass index and physical activity. Which variables and their interaction terms are statistically associated with diabetes and only include Interaction terms which are appropriate? Choose your final model between enter and forward LR method and then perform model diagnostics. How do you interpret the effect size and 95% CI of your final model?

Answer 1

a. Null hypothesis :it is a hypothesis used in clinical studies in which the observed results are not matching with the expected results, we call it as null hypothesis.

Alternate hypothesis :when the null hypothesis is rejected, we will accept the actual results that are obtained during the study, this is called alternate hypothesis.

b. The study included many variables like blood glucose, physical activity, smoking, alcohol consumption, BMI etc. If we assess the relationship between BMI and glucose, it is not only related to glucose, but also to other variables like physical activities, diabetes etc. Hence the study will have wide range of variables even if we consider one variable at a time.

c. The significance of the study will depends on p value in randomised controlled trial and Odds ratio in case control study. After collecting all the data, we have to find out p value, if it is less than 0.01 ,then the study is significant.