Topic 9: measure of association of 2 numeric variables Flashcards
(10 cards)
1
Q
Describe association between numeric variables
A
- Don’t use mean
- Use mathematical model to predict change in outcome for change in exposure
- Quantify strength of association = correlation coefficient
2
Q
3 steps to investigate association between variables
A
1) Derive scatter plot
2) Perform correlation analysis
3) Perform linear regression analysis
3
Q
Explain a scatterplot
A
- Exposure = independent = X
- Outcome = dependant = Y
- Provide overall impression for association between variables
- Reveal trend = direct positive/inverse negative association
- Direct positive = X increases Y increases
- Inverse negative = X increases Y decreases
4
Q
Define correlation
A
- Association between numeric variables
- Positive/negative
5
Q
Define correlation coefficient
A
- Strength of correlation between variables
- Ranges from -1>+1
- +/- = direction of association
- 0 = no correlation
- > 0.7 = strong
- 0.5-0.7 = moderate
- 0.3-0.5 = weak
- < 0.3 = very weak
6
Q
Define Pearson’s correlation
A
- Most used
- Correlation between 2 variables using original values
- Used = continuous data + normal distribution + linear relationship
7
Q
Define Spearman’s correlation
A
- Correlation between 2 variables by 1st ranking values then assessing correlation between ranks
- Used = ordinal data + not normal distribution + non-linear relationship
8
Q
Describe linear regression
A
- Asseses extent to which increase in 1 variable is associated with increase in another variable
- Line of best fit by using least square method
9
Q
When is it line of best fit?
A
- Overall difference between actual values is at minimum
10
Q
Formula for line of best fit
A
Y’ = a + bX
- Y’ = predicted value for Y
- b = slope = regression coefficiant
- Regression coefficiant = estimates change in Y for each 1 unit increase in X
