Epidemiology-Regression and Correlation

Question 1

What is the relationship between correlation and regression?

Answer 1

Correlation is included under the umbrella of regression. Regression is a family of methods for relating a predictor to an outcome. Correlation of a measure of the strength of the association between two variables.

Question 2

What determines the strength of association between two variables?

Answer 2

Correlation

Question 3

What predicts one variable from another?

Answer 3

Regression

Question 4

What allows you to estimate the parameters of a predictive model?

Answer 4

Regression

Question 5

How do you interpret a correlation coefficient?

Answer 5

0= straight line, 1= lower left to upper right diagonal line and -1 = upper left to lower right diagonal line

Question 6

How do you test to see if there is a statistically significant correlation between X and Y?

Answer 6

T test

Question 7

When do you use a standard Pearson correlation (p-value)?

Answer 7

When the sample is random and at least one variable follows a normal distribution

Question 8

What method do you use to test for statistical significance of a correlation if you cannot use a p value?

Answer 8

Spearman rank correlation. The ranking system gives outliers less influence.

Question 9

What does a regression line do?

Answer 9

It minimizes the sum of the squared distances from the average line of the data.

Question 10

What formula do you use to predict values from a regression line?

Answer 10

y=mx+b

Question 11

How can you deal with confounding?

Answer 11

Stratification, matching and adjustment (via multiple regression which can add a third predictor)

Question 12

What formula do you use to predict values from a multiple regression line?

Answer 12

y = m(1)x(1) + m(2)x(2) + b

Question 13

How does logistic regression estimate relationships between variables?

Answer 13

It estimates odds ratio between categorical variables (not continuous variables).

Question 14

How would you assess association between two continuous variables as seen below?

Answer 14

Notice a spot with really high prevalence and spot with negative prevalence. This won’t work. You can analyze the log of the odds ratio of a continuous predictor and a categorical outcome.

Question 15

How do you get the odds ratio from a log odds ratio graph?

Answer 15

Take the anti log

Question 16

BMI is a continuous value. How do you interpret the odds ratio from a continuous predictor and a categorical outcome?

Answer 16

The odds of the categorical outcome a “x” higher/lower with each additional BMI unit.

Question 17

When should you use logistic vs. linear regression?

Answer 17