Dovid Moreau Module Flashcards
(11 cards)
What does a Bayes rule allow us to figure out in a research setting?
Bayes Rule allows us to figure out the probability of A happening given B happens. Used in medical testing, helps us figure out false positive and false negative chances.
Use prior knowledge and data to make probabilistic inferences
If you change the prior how does the graph change? If you changed the observed data, what might change on the graph?
Posterior will be between the prior and observed data
larger sample size will cause posterior and likelihood to be closer
When you look at a graph showing the prior, likelihood, and posterior, what does each curve tell you? What is on the x and y axis of the graph?
Prior - Hypothesis
Likelihood - Observed data
Posterior - Updated belief after seeing data - in the middle
X axis - Parameter values e.g.
Y-axis - Probability density
What does a Bayes Factor tell us? What do we conclude if the BF is very large? Very small? Exactly 1?
BF tells us ratio of evidence for one H versus another
BF < 1 means bottom hypothesis more likely than top
BF > 1 means top hypothesis more likely than bottom
BF = 1 means hypotheses are equally likely
What is sensitivity?
Probability that a test will be positive among those with the disease
What is specificity?
The fraction of those without the disease who will have a negative test result
How do you work out the probability of having a disease given a positive test?
Steps to Calculate
Identify:
Sensitivity = true positive rate
Specificity = true negative rate
Prevalence = % of population with disease
Plug into Bayes’ Theorem:
PPV= Sensitivity×Prevalence/
Sensitivity x Prevalence + (1-Specificity) x (1-Prevalence)
Example
Sensitivity = 0.9
Specificity = 0.95
Prevalence = 0.01
PPV = 0.9 x 0.01/
0.9 x 0.01 + (1-0.95) x (1-0.01)
What are the core differences between Bayesian and frequentist methods?
Bayesian:
- Changing
What are p-hacking and HARKing? What are some of the problems with questionable research practices?
P-Hacking - the misuse of data analysis to find statistically significant results - even if those results are not meaningful or representative of a true effect
HARKing - Hypothesising after results are known
How do questionable research practices affect the scientific literature? What is preregistration and how does it help?
They make any findings less valid and harder to draw wider findings from. Meta-analysis dont work
Pre-registration ensures that we are aware what studies are occurring and the hypothesis for them are. Can adjust based off that, even if the findings of the studies aren’t published. Encourages researchers to publish results they normally wouldn’t (file-drawer problem).
What is prevalence
The true portion of the population with the disease
