Stats/ critical numbers Flashcards
what is a case control study
Find people with disease and controls look back in time for exposure
what are the pros of case control
- Good for rare outcomes
- Fast cheap
- Few ethical considerations
cons of case control
- cannot prove causation/ eliminate confounding factors
- difficult to establish event order
- subject to recall bias
- only investigate single disease
Describe cross sectional study
Sample a population and count number of affected and unaffected people
pros of cross sectional
generates hypothesis
cheap fast
few ethical considerations
cons of cross sectional
cannot prove causation/eliminate confounders
less suitable for rare disease
doesn’t establish event order
sample bias could occur
what the cohort study do?
take a population monitor those with and without exposure for linked outcome
prospective
pros of cohort study
event sequence clarity
few ethical considerations
cons of a cohort study
Cannot rule out confounding factors not suitable for rare disease time consuming and expensive requires follow up - could be difficult patients can change behaviors
Wat do be a randomised controlled trial
multiple group organised into arms given different exposures/treatments and comparing outcomes.
Arms can be balanced by matching, randomising, cross-over, placebo, and blinding.
pros of RCT
gold standard in proving causation and eliminating confounding factors and bias
cons of RCT
expensive time consuming
not good for rare diseases
ethical approval trickier as gold standards are often unethical
requires compliance
what is ecological study?
using massive samples to look at trends across populations not individuals.
pros of ecological
fast cheap
large sample
easy
good first step for hypothesis generation
cons of ecological
don’t know how data was collected
data may be missing
doesn’t prove causation
correlation not causation - ecology fallacy
what are the types of sampling
random systematic convenience cluster stratified
lead time bias
illness detected sooner but no difference in treatment outcome
length time bias
people who live with disease for longer more probable to be screened - false impression of screening improving survival
what is risk
risk is probability
eg risk of rolling a 6 is 1/6
how do you calculate number needed to treat/harm (NNT/H)
1/ absolute risk difference
must always be rounded up
how to calculate odds
x/(n-x)
x=outcome
n=sample size
with/without
how to calculate sensitivity
true positive/all population with disease
true pos/(true pos + false pos)
how to calculate specificity
false positive/all healthy patients
think who is healthy but have been told they’re ill
positive and negative predictive value
disease pos/all who’s results were pos
disease neg/ all who tested neg
test accuracy how to calc
all correct results/ total tests performed
all results
standard deviation equation
root of [sum of (each value-mean)^2 / n-1]
standard error is?
standard deviation / n^(1/2) [which is square root n]
when is interquartile range used and how is it calculated?
when data is skewed
upper quartile - lower quartile
Q3-Q1
what is a 95% confidence interval
a range we are 95% sure the true population lies in
how to calc 95% confidence interval?
mean +or- 1.96*SE
what is correlation?
simple association - no causation
when to use regression
when a change in one variable can predict another
what is a test for correlation
pearson’s correlation coefficient
1 pos
-1 neg
0 nil
when is a logistic regression best used
outcome is binary
log(outcome)= a + bx
a is where prob increases
b is how fast it increases
Carried out in a snap-shot of time without follow-up of subjects or looking back in time
Cross Sectional
Collect information now and follow subjects up over time to explore outcomes
Cohort
Collect information on an outcome now, and look back in time to see what exposures were experienced
Case-Control
Information on groups of individuals (e.g. countries) rather than individual level data
Ecological
