week 6 Flashcards
ratio
division of 2 frequencies
one quantity not a subset of the other
event rate/risk
probability that the even will occur
proportion
freq in one group / total
chi square tests
data are categorical
numbers are frequencies
mutually exclusive and exhaustive
2 types
goodness of fit: tests observed proportions against a known proportion
test of independence: tests for an association between the 2 variables
need a p value to determine if there is an association or not
relative risk
how many times more or less likely a person exposed to a condition develops the condition compared to an unexposed person
RR=(incidence of outcome with exposure)/(incidence of oucome without exposure)
if RR=1, not likely
if RR>1 more likely in exposed
if RR
odds ratio
how many more times likely the odds of finding an exposure in someone with a disease is compared to finding an exposure in someone without the disease
OR=(odds of exposure in those with disease)/(odds of exposure without disease)
confidence intervals
range of possible values around the point estimate
95% probability that the true value falls within the interval
width determines precision
may be used for significance (if 1 in included in RR interval, not significant)
confounder
a third variable that influences the exposure/treatment-outcome relationship
multivariate analysis in experimental research
adjustment for confounding is needed when randomization isnt successful
multivariate analysis in observational research
adjustment for confounding is needed because its not randomized
linear regression
used for prediction of an outcome based on an independent variable
best fit lines
numeric variables
logistic regression
used for dichotomous outcomes
r squared value
indicated how much variability in y is explained by the model
association vs effect
association implies a mere statistical relationship
effect implies that the relationship is causal
