Quiz 4

Question 1

Linear regression vs logistic regression

Answer 1

Linear regression assesses strength of relationship between 2+ continuous variables

Logistic regression produces an odds ratio for a categorical dependent variable and a continuous independent variable

Question 2

Uses of linear regression

Answer 2

Can be used for predictions (regression line attempts to predict relationship) and for identifying confounders through comparison of simple and multiple linear regression

Question 3

What is the null hypothesis in linear regression?

Answer 3

Null hypothesis is that B1 slope = 0
P value < 0.05 rejects H0 and signifies significance of association

Question 4

Multilinear regression answers what question?

Answer 4

Which independent variables are predictors of the dependent variable?

Question 5

r value vs R^2 value

Answer 5

r value = strength of correlation (0-1) between 2 independent variables

R^2 = coefficient of determination, the variance of dependent variable which can be explained by independent variable which reflects overall model and is given as percent

Question 6

Linear regression outputs

Answer 6

Model summary: R^2 value

ANOVA table: sum of squares, df, mean square, F and p value

Parameter estimates (coefficients): unstandardized, standardized

Question 7

Assumptions for linear regression (5)

Answer 7

Residuals are normally distributed
Linear relationships between variables in question
No extreme outliers
Minimum 10 cases per independent variable
No multicollinearity between independent variables (r > 0.7)

Question 8

What are confounders?

Answer 8

Confounders are variables related to both exposure and outcome which distorts (stronger or weaker) estimate of predictor-outcome association

Visualized: directed acyclic graphs (DAGs)

Can be controlled if measured during data collection

How to identify a confounder: conduct linear regression with and without confounder → if p value was significant without confounder or β coefficient changes >10% → it is a confounder

Question 9

Uses for logistic regression

Answer 9

case control, cross-sectional, and prospective cohort studies to measure prevalence at one time point

Question 10

Odds ratio vs relative risk in logistic regression

Answer 10

Odds ratio: odds that exposed has outcome / odds that unexposed has outcome
(a/b)/(c/d)

Relative risk: incidence of outcome among exposed / incidence of out among unexposed
[a/(a + b)] / [c/(c+d)]

Question 11

SPSS output for logistic regression

Answer 11

Exp(B) odds ratio of outcome

B value less relevant

Question 12

Logistic regression modelling for covariate assessment

Answer 12

Either:

Only include significant predictors

Or choose covariates based on literature whether they are significant predictors or not (limits hypothesis testing)