Stats Flashcards
Cohort study
Classify by EXPOSURE STATUS and follow for disease/outcome
prospective or retrospective
- if prospective- allows calculation of incidence of disease
ideal when exposure is rare
inefficient for studying rare diseases
case-control study
study groups defined by DISEASE STATUS
good for diseases with long latency periods
good for rare diseases
bias
inefficenct for rare exposures
unable to calculate incidence rates
phases of a clinical trial
I: intervention tested on a small cohort to evaluate safety and dosing
II: intervention tested on a larger cohort to determine effectiveness
III: intervention tested on an even larger cohort- effectiveness and side effects, compare to other treatment modalities
IV: after release of intervention/drug to establish effectivness/side effects
cross-sectional study
descriptive
exposure and disease status measured together at one point in time
requires large sample size
disease PREVALENCE
relative risk
estimates the magnitude of an association between exposure and disease
indicates the likelihood of developing the disease in the exposed group relative to the non-exposed group
RR = 1.0: no association between exposure and disease
RR > 1.0: positive association/increased risk of disease given exposure
RR < 1.0: negative association/decreased risk of disease given exposure
AR in exposed/AR in non-exposed
[a / (a+b)] / [c / (c+d)]
odds ratio
gives odds of disease in exposed over odds of disease in non-exposed
measures strength of association
cross product ratio = a x d/ b x c
rare disease: OR very close/approximates RR
often used with case-control studies
number needed to treat
NNT
a way to measure the benefit of an intervention
NNT = 1/AR
**AR = [c/(c+d)] - [a/(a+b)]*
sensitivity
% of true disease correctly identified by screening test
probability of test being positive when true disease is present
a / (a + c)
specificity
% of population free of disease correctly identified as not having the disease by screening test
probability of test negative when true disease is absent
d / (b + d)
positive predictive value
PPV
% of individuals who tested positive that do have true disease
probability of having true disease when test is positive
a / (a+b)
negative predictive value
NPV
% of individuals who tested negative that do NOT have true disease
probability of not having true disease when test is negative
d / (c+d)
null hypothesis
NO ASSOCIATION between 2 variables
one tail vs two tail
one tail: asses change in ONE direction
two tail: utilized in EITHER direction
confidence interval
range of values that is intended to contain the parameter of interest with a certain degree of confidence (95%, 99%)
the larger the level of confidence, the larger the interval
as sample size increases, width of the confidence interval decreases
if interval crosses 1 (= NO ASSOCIATION), the association being tested in not considered statistically significant
null hypothesis cannot be rejected
parametric
NORMAL DISTRIBUTION = BELL SHAPED CURVE
type 1 error
ALPHA ERROR
REJECT NULL HYPOTHESIS WHEN IT IS TRUE
ex: say there is an association when there isnt
FALSE POSITIVE
type II error
BETA ERROR
DO NOT REJECT NULL when null is fase
FALSE NEGATIVE
ex no difference but there is actually a difference
usually due to inadequate sample size
power
1- type II error (beta)
to increase power:
increase alpha value
increase sample size
increase magnitube of difference between the two populations
prematurity rate
< 37 weeks
fetal mortality
early < 20 weeks
intermediate = 20-27 weeks
late >/= 28 weeks
perinatal mortality
fetal > 28 weeks + neonatal < 7 days
neonatal mortality
< 28 days
post neonatal
28-364 days
infant
< 1 year
birth weight definition
normal >/= 2500g
low birthweight < 2500g
very low birthweight < 1500g
extremely low birthweight < 1000 g
memorize this
types of data
mode–>median–>mean
