Evidence Based Medicine Flashcards
What is the summary of a hypothesis test?
1) Start by saying there is no association (null hypothesis)
2) Compare what you observed with the null hypothesis
3) Work out the probability that the difference between observed and expected happened by chance (P value)
What is a P value?
How do we interpret a P value?
The probability that the observed association happened by chance
Small P value = < 0.05 (significant) - reject null hypothesis, there is association
Large P value = > 0.05 (not significant) - accept null hypothesis, there is no association
What statistical test would you use for 2 sets of categorical data where the observations are not paired?
Chi-squared test
What statistical test would you use for continuous data where the observations are normally distributed and independent?
Independent samples T-test
What is the purpose of an independent samples T-test?
To calculate the difference between means, standard error of difference and calculate T to compare against values from T-distribution
Give an example where Independent samples t-test would be used
Investigating change in blood pressure after statins given
Give an example where we would use a paired T-test
Measure lung function, give drug, measure again
What can we do if data is not normally distributed?
Try log transforming data or use a non-parametric test (Mann-Whitney)
What is the regression coefficient(β)?
An estimate of how much y increases/decreases for each unit increase in x
y = α + βx
α = intercept β = gradient
What are the differences between a positive and negative β coefficient?
Positive = outcome increases as exposure increases Negative = outcome decreases as exposure increases 0 = null hypothesis
What is correlation?
The degree of linear association between 2 variables. Correlation coefficients lie between +1 and -1
What is ANOVA (ANalysis Of VAriance)
Extension to t test that compares means in >2 groups
Why is it important to adjust for cofounders?
Because they can interfere with odds ratios and lead to different results
What does an r squared value show us?
How much of the outcome can be accounted for by the model
What is the difference between at type 1 and type 2 error in hypothesis testing?
Type 1 error - false rejection of a true null hypothesis
Type 2 error - failure to reject a false null hypothesis (inadequate sample size?)
