10th Feb

Question 1

General linear modelling (GLM)

Answer 1

Helps to indicate if there is an association between 2 variables

Question 2

Nuisance variables

Answer 2

Confounders and competing exposures - can undermine the interpretation of these associations - we can adjust for these

Question 3

Stata command regress

Answer 3

Used to build linear models

eg regress weight height (predict weight from height)

Question 4

Stata command regress with categorical variables

Answer 4

Put xi before regress
Indicate which variables are categorical by putting i in front
eg
xi: regress weight i.sex

Question 5

Logistic regression

Answer 5

When outcome is categorical eg did attend/ did not attend

eg logistic complynot age

Question 6

R2

Answer 6

Proportion of variation explained by linear model is R2
R2 is a value between 0 and 1
Higher R2 values indicate that more of your variation is explained by your model
1 is a perfect fit (your model perfectly fits the outcome)

Question 7

R2 and logistic regression

Answer 7

Fit is reported by psuedo-R2 as standard R2 cannot be calculated for logistic regression