linear regression

Question 1

Why Regression?

Answer 1

• Regression tells us more about our data than correlation
• It’s the foundation to other types of regression analysis (some of which you will cover in your course)
• Linear regression used to test how well one variable predicts another variable

Question 2

Correlations tell us:

Answer 2

• strength of relationship between two continuous variables

* statistical significance

Question 3

Linear Regressions tell us:

Answer 3

• strength of relationship between two continuous variables
• how much one variable changes as another variable changes
• the value of one variable if the other variable was 0
• can predict a person’s score on a variable
• statistical significance

Question 4

Variability in data (revision)

Answer 4

• If we ask people’s RPE during a 5km run, there will be lots of variability
• As Psychologists, we want to know why

Question 5

Building a model

Answer 5

Caffeine consumption ——————————– RPE

Causality Caution
Regression is used to predict an outcome variable
BUT
We still can’t infer causality

Question 6

Effect Size

Answer 6

• R2
• strength of relationship between two continuous variables
• % variance explained by the model
• Deviation from each data point and the line of best fit

Question 7

slope

Answer 7

• Beta (𝛽)
• How much one variable changes as another variable changes
• Slope of the line of best fit

Question 8

intercept

Answer 8

• a
• the value of one variable if the other variable was 0
• Where the line of best fit intersects the x-axis

Question 9

predicting values

Answer 9

• a
• the value of one variable if the other variable was 0
• Where the line of best fit intersects the x-axis

Question 10

statistical significance

Answer 10

• p
• Likelihood of observing this effect if there’s no real effect in the population
• p < .05 means there is less than 5% likelihood that our results would not be found in the population (i.e. there’s only a small likelihood!)

Question 11

summary

Answer 11

• Linear regression used to test how well one variable predictors another in our sample
• It also tells us:
• the strength of the relationship ( 𝑅﷮2﷯)
• how much the outcome variable changes for each increase in one of the predictor variable (𝛽)
• and can predict values of the outcome variable if we know their value of the predictor variable
• Significance testing tells us the likelihood of finding this effect in our sample if there’s no effect in the population