You will be using the “Heart” dataset for this lab. You will be using the variables Sex, ChestPain and HeartDisease.
/content/enforced/194732-2211FS0523411A00/heart.omv
You will be building Log-Linear models using this data.
First, check the variables to make sure they are set to be the correct variable type. Make sure to get in the habit of doing this
Second, look at the Descriptives of these three variables and look at the Frequency tables. You will be re[porting some of these values. It always good practice to do this.
Third, build a Log-Linear model with Sex and HeartDisease. Make sure the model includes the two main effects and the interaction term in it. Remember, to build the interaction term you may need to hold shift and select the two predictors and then drag them over into your Block to add the interaction.
Your research question is “Does being female decreases the odds of having heart disease as compared to being male?” You need to set the reference levels to address this question.
Recall what the interaction term is. It is the LOG of the reference cell. So to get back to the COUNTS of the reference level calculate: e^(value). To confirm that you are doing things correctly, check the following. Go to Estimated Marginal Means in the analyses setup and drag BOTH sex and HeartDisease into “Term 1.” Make sure that Marginal means tables is clicked. Then look at this table and the value in the cell you WANT to be the reference cell should be the same value as what you calculated as e^(intercerpt). Also remember, we know that females have less heart disease than males; therefore, you would expect that the interaction term would be negative to reflect the lower value of heart disease and the direction of your research question.
USE THIS FIRST MODEL TO ANSWER QUESTIONS 1 – 6.
=====================================================
Now add Fasting blood sugar to the model (fbs), 0 for normal and 1 for elevated. Add this variable as a main effect AND add all interaction terms. Your model will have 7 things in it: 3 main effects, 3 two way interactions and 1 three way interaction.
FOR QUESTIONS 7 – 10 USE THE FOLLOWING RESEARCH QUESTION
The research question is “Does elevated fasting blood sugar predict a difference in the odds of having heart disease more in males than females?”
You will be using the “Heart” dataset for this lab. You will be using the variables Sex, ChestPain and HeartDisease.
/content/enforced/194732-2211FS0523411A00/heart.omv
You will be building Log-Linear models using this data.
First, check the variables to make sure they are set to be the correct variable type. Make sure to get in the habit of doing this
Second, look at the Descriptives of these three variables and look at the Frequency tables. You will be re[porting some of these values. It always good practice to do this.
Third, build a Log-Linear model with Sex and HeartDisease. Make sure the model includes the two main effects and the interaction term in it. Remember, to build the interaction term you may need to hold shift and select the two predictors and then drag them over into your Block to add the interaction.
Your research question is “Does being female decreases the odds of having heart disease as compared to being male?” You need to set the reference levels to address this question.
Recall what the interaction term is. It is the LOG of the reference cell. So to get back to the COUNTS of the reference level calculate: e^(value). To confirm that you are doing things correctly, check the following. Go to Estimated Marginal Means in the analyses setup and drag BOTH sex and HeartDisease into “Term 1.” Make sure that Marginal means tables is clicked. Then look at this table and the value in the cell you WANT to be the reference cell should be the same value as what you calculated as e^(intercerpt). Also remember, we know that females have less heart disease than males; therefore, you would expect that the interaction term would be negative to reflect the lower value of heart disease and the direction of your research question.
USE THIS FIRST MODEL TO ANSWER QUESTIONS 1 – 6.
=====================================================
Now add Fasting blood sugar to the model (fbs), 0 for normal and 1 for elevated. Add this variable as a main effect AND add all interaction terms. Your model will have 7 things in it: 3 main effects, 3 two way interactions and 1 three way interaction.
FOR QUESTIONS 7 – 10 USE THE FOLLOWING RESEARCH QUESTION
The research question is “Does elevated fasting blood sugar predict a difference in the odds of having heart disease more in males than females?”
Questions:
What is the difference in counts between the reference level cell and the cell the interaction term refers to? Look at the marginal means to figure this out.
1) How many people have heart disease?
2) How many females have heart disease?
3) What is the intercept parameter estimate for the model with Sex and Heart Disease in it?
4) How many counts does this correspond to?
5) Does this data support your research question? (Y or N)
6) Does this data support the DIRECTION of your research question? (Y or N)
Now add fasting blood sugar to the model (fbs). Make sure that the three main effects and all of the interactions are included in the model.
7) Which sex is the reference level? (M or F)
8) Which level is the reference level for fbs? (0 or 1)
9) What is the parameter estimate for the three way interaction term?
10) What is the difference in counts between the reference level cell and the cell the interaction term refers to? Look at the marginal means to figure this out.
Note, I would be a bit sceptical of this significant result due to the small sample sizes in some of the cells. It is important to point out that even with large sample sizes when you break data down into more and more categories you can end in small cell counts.
