regression

Question 1

why would you carry out a regression?

Answer 1

believe or predict that there is a causality

Question 2

if you had two outcomes what test would you do?

Answer 2

binary outcome - logistic regression

Question 3

what are the tests for continuous outcomes with one or more predictors?

Answer 3

• 1 predictor - simple linear regression

- >1 predictor - multiple linear regression

Question 4

what is the method of least squares?

Answer 4

• a method to plot the line of best fit
• it the line that minimises the sum of squares of the distances between the observed values of the response; and the values predicted by the model

Question 5

what does e1 refer to?

Answer 5

residual error - variation around the line

Question 6

How does sums of squares apply to the model?

Answer 6

SSt - total sum of squares - amount of difference when the most basic model is applied
SSr - residual sums of squares - degree of inaccuracy when the best model is fitted to the data - how much variability cannot be explained by the model
SSm - how much variability can be explained by the model we fit to the data - how much better is regression line compared to mean fit

Question 7

What dose the f ratio determine?

Answer 7

how much variability can be explained by the model (SSm) compared to how much variability cannot be explained by the model (SSr)

Question 8

what is the correlation of determination (R2)?

Answer 8

• how much variability can be explained by the model (SSm) compared to how much variability there was in the first place (SSt)
• the percentage of variation in the dependent variable that can be accounted for by the regression model variables

Question 9

for multiple linear regression what are some methods of selecting variables to keep in the model?

Answer 9

• IBM SPSS
• forward
• backward
• stepwise
• enter
• think about if the variable is a significant predictor
• does the model fit improve with addition/removal of variables

Question 10

What criteria test to see how good the model/line is?

Answer 10

• F-ratio
• Correlation of determination
• Hypothesis testing of model coefficients

Question 11

What does the hypothesis testing of model coefficients do?

Answer 11

• tests two null hypotheses
• the intercept is equal to zero
• the gradient of the slope is equal to zero