Critical Numbers Flashcards
Name the 5 types of study design
Case control
Cross-sectional
Cohort
RCT
Ecological
What are case control studies?
Retrospective studies that find people with an outcome + look back and see if they have the risk factor in question
Name 2 positives of case control studies
Fast + cheap
Good for rare outcomes
Name 2 negatives of case control studies
Can’t prove causation/can’t eliminate confounders
Difficult to establish order of events
What do cross-sectional studies look at?
An outcome at one point in time
Name 2 positives of cross sectional studies
Fast + cheap
Generates hypotheses
Name 2 negatives of cross-sectional studies
Not causative
Less suitable for rare diseases
What do cohort studies do?
Prospective follow-up of an outcome over time, have separate groups, some with , some without exposure
Name 2 positives of cohort studies
Little ethics
Can clearly sequence events
Name 2 negatives of cohort studies
Can’t prove causation
Time-consuming + expensive
What are RCT’s?
Gold standard study, has multiple groups which have different exposures + compare outcomes. Can use crossing over, placebos, double-blind matching etc
Name 2 positives of RCTs
Can prove causation
Random=less bias
Name 2 negatives of RCTs
Time consuming + expensive
Often unethical
What are ecological studies
Large population-look at data, prevalence, correlation + trends
Name 2 positives of ecological studies
Fast + cheap
Name a negative of ecological studies
Ecological fallacy-correlation does not always equal causation
Name the 5 types of sampling
Random
Systematic
Quota
Cluster
Stratified sampling
Give an example of random sampling
Using a random number generator
Give an example of systematic sampling
Numbering people 1,2,3,4 and assigning them to one of four groups
Give an example of quota sampling
Filling up a group until you have all the people you need
Give an example of cluster sampling
Divide people into groups (usually geographical), each group=1 cluster, clusters randomly selected
Give an example of stratified sampling
Divide into male/female, then take random sample of each
What is sample bias?
Non-representative sample
What is recall bias?
Lack of specifics/can’t remember details
What is social desirability?
Lying about things for societal reasons
What is information bias?
Intentionally wrong data measure
What is volunteer bias?
Volunteers unlikely to be representative
What is selection bias?
Similar to volunteer bias-choosing of volunteers isn’t random
What is lead time bias?
Screening seems to improve life expectancy but actually just picked up earlier
What is length-time bias?
Those with a less aggressive condition/longer living are more likely to be picked up
What are confounders?
Factors related to both the outcome + exposure that may affect results + need to be accounted for
What does PICO stand for?
P-patient/population
I-intervention
C-control/comparison
O-outcome
Name 3 types of variable
Binary
Categorical
Numerical
What are binary variables?
Only two answers e.g. yes/no, true/false
What can categorical variables be divided into?
Nominal
Ordinal
What are nominal variables?
They have no distinct hierarchy e.g. hair colours
What are ordinal variables?
They have a hierarchy e.g. social classes
What can numerical variables be divided into?
Continuous + discrete
What are continuous variables?
Numbers that can take any value within a range e.g. height, age
What are discrete values?
Whole numbers e.g. number of people
What is standard deviation?
Measures spread of data around the mean for this population
What is the standard error?
Measures spread of several means from various populations + estimates how far away from sample mean the true mean is
What is the difference between the SD + SE?
SD=spread around mean, SE=estimates real mean
How do you calculate standard error?
Sample SD/ square root of number of samples
What is kurtosis?
Vertical data skew
What is the interquartile range + how do you calculate it?
Measure of spread, UQ-LQ
What is the confidence interval?
Range of values the population mean is likely to be within e.g. 95% CI
How do you calculate the risk difference?
Risk-risk
How do you calculate risk?
Number of examined outcomes/total possible
How do you calculate risk ratio(/relative risk)?
Risk/risk
What should the relative risk be lower than for it to be a low risk?
Less than 1
How do you calculate odds ratio?
Ratio of (probability of event): (probability of non event)
How would you write a null hypothesis?
There will be no difference…
How would you write an alternative hypothesis?
There will be a significant difference…
What does the P value show?
Probability result is due to chance
What is the equation for simple linear regression?
y=mx+c
What does multiple linear regression do?
Considers several independent variables/confounders
When is logistic regression used?
When an outcome variable is a binary variable
When do you use multiple regression as opposed to correlation?
Correlation=no distinction between variables, no causation, simply association
Multiple regression=y causes x, prediction
What do you use scatter plots to show?
Correlation
What do regression models do?
Try to predict data using an equation
What do the letters stand for in y=mx + c?
y=dependent variable
m=gradient
x=independent variable
c=y-intercept