Stats Flashcards
What do screening tests rely on
What is it a measure of
Good sensitivity- true positives
What measures true negatives
Specificity
How do you calculate PPV and NPV
PPV- true +/ (true+ + false +)
NPV- true -/ (true- +false -)
What is a type 1 error
What does it give
False positives
P values- normally 0.05%
What is a type 2 error
What makes it more likely
False negatives
Of the study is underpowered
Which variable won’t change no matter how sick the individuals are
Relative risk
How do you calculate relative risk
A/a+b divided by c/c+d
How do you calculate odds ratio.
Abxcd
How do you calculate absolute risk reduction
What does this allow you to calculate? Why?
Events in control/total control - events in exposure/total exposure
NNT 100/(absolute risk reduction in %)
What do the following phases of drug trials focus on 0- 1- 2- 3- 4-
0- pharmacokinetics and pharmacodynamics
1- safety
2- dosing requirements and some efficacy
3- confirms efficacy- usually against standard of care (RCTs)
4- post marketing
What does the Bradford hill criteria assess
What does it not take into account
Evidence for a casual relationship
Sensitivity
What are 2 types of experimental studies
Control trials and RCTs
What are 2 examples of observational studies
How do they differ
Cohort study- looks prospectively
Case control- looks retrospectively
What does an ecological study look for
Variables within a group
What is an example of a cross sectional study
Community survey
What does selection bias entail? Recall? Confirmation? Confounding?
Which type of study is worst for recall bias
Not randomising
Not following up everyone- worse with case control
Ignoring evidence against your fixed beliefs
Mixing of effects
How do you calculate a positive likelihood ratio
“Negative
Sens / (100-spec)
100-sens/specificity
Forest plots What do they represent What does the solid vertical line mean What do the boxes mean? What does a big vs small box mean. What do the horizontal lines mean What does the diamond mean
Visual representation of a meta analysis
Line of no effect
Study outcomes. Bigger box means bigger sample size
Confidence intervals of each study
Overall studies. Tips=95% CI
What would a highly significant study favouring treatment look like on a forest plot
Diamond far to the right of the line of no effect
Forest plot
If the horizontal line crosses the line of no effect what does this mean
The study is not statistically significant!
Does sensitivity rule out or in
Rules out!
