LEC 12 Simple & Multiple Logistic Regression

Question 1

When to do logistic regression?

Answer 1

Nominal (dichotomous) variables (dependent variable)

Question 2

2 types of logistic regressions

Answer 2

1. Simple logistic regression
   - 1 independent variable
2. Multiple logistic regression
   - >=2 independent variables

Question 3

Simple logistic regression model equation

& Odds ratio equation

Answer 3

loge(odds of outcome) = alpha + beta(x)

odds ratio (OR) = e^beta

Question 4

Multiple logistic regression model equation

& Odds ratio equation

Answer 4

loge(odds of outcome) = alpha + beta(xi) + …

odds ratio (OR) = e^beta(i) after controlling for all other independent variables

Question 5

Range of values of OR

Answer 5

0 - infinity

Question 6

Simple logistic regression

- to test null hypothesis __

Answer 6

To test the null hypothesis that there is no association between the independent variable and the dependent variable

Question 7

Multiple logistic regression

- to test null hypothesis __

Answer 7

To test the null hypothesis that there is no association between the independent variable(xi) and the dependent variable, after controlling for all other independent variables

Question 8

Odds ratio calculation for cross-product ratio

Answer 8

= (no. of events or outcome)/(no. of no events or outcome)
= ad/bc

no. of events or outcome = a/c
no. of no events or outcome = b/d

Question 9

Simple logistic regression assumptions (3)

Answer 9

1. The dependent variable should be a dichotomous variable (2 categories only)
2. The observations are independent of one another
3. There is a linear relationship between the independent variable and the loge(odds of outcome)
   - scatter plot to predict

Question 10

Multiple logistic regression assumptions (4)

Answer 10

1. The dependent variable should be a dichotomous variable (2 categories only)
2. The observations are independent of one another
3. There is a linear relationship between the independent variable and the loge(odds of outcome)
   - scatter plot to predict
4. There is little or no multicollinearity among the independent variables / independent variables should not be too highly correlated with each other
   eg weight and BMI

Question 11

OR = 1

Answer 11

• no association between exposure and outcome

Question 12

OR >1

Answer 12

• positive association between exposure and outcome (_ times more)

Question 13

OR<1

Answer 13

• inverse association between exposure and outcome (_% reduction)

Question 14

Simple logistic regression hypothesis

Answer 14

Ho : OR = 1

H1 : OR =/ 1 or OR<1 or OR>1

Question 15

Multiple logistic regression hypothesis

Answer 15

Ho : ORi = 1

H1 : ORi =/ 1 or ORi<1 or ORi>1