Biostatistics Flashcards
Specificity and Sensitivity of variant tests
highest sensivity at beginning of disease curve (point A)
highest specificity at end of healthy curve (point E)
Alpha value
probability of type I error (rejecting a true null hypothesis)
P-value
probability of obtaining a result as extreme as the given result while assuming the null hypothesis is true
indicates statistical significance
Confidence intervals
chance that the difference between two groups reflects a true difference
Case fatality rate
number of fatalities due to specific illness among all individuals with that illness
Standard deviation
2 standard deviations = 95% if normal distribution
Odds ratio
AxD/BxC
(A/B)/(C/D)
ANOVA
compare means of 2 independent groups. will yield p-value but more difficult than t-test
Chi-square test
analyze non-continuous data
paired t-test
compare means of samples that are correlated (non-independent)
2-sample t-test
analyze continuous data from 2 separate samples with normal distribution
1-sample t-test
compare mean of single group to particular value (but not 2 groups)
Recall bias
occurs in retrospective studies where participants are asked to recall past risk factors
Performance bias
treatment that is administered differently at multiple sites
lead-time bias
increased survival time due to earlier detection of disease (without a change in disease course)
Skew on distributions
Positive skew : meAn > meDian > mOde
What increases power of study?
increase sample size or increase effect size
Cohort study
observational study , expansion of case series or cross sectional
provide estimate of incidence of a disease, outcome, or prognosis
strengthens by dissociating exposure from outcome
inefficient for diseases of long latency
Accuracy
calculated by true positives and true negatives / total
Case control study
usually retrospective
compares patients w/ dz to those w/o dz
look at past data to determine risk factors or exposure
Cross-sectional study
observational study measures risk factor and outcome
compares different population groups at a moment in time
can compare multiple variables
does not determine causation
Sampling bias
occurs in studies that recruit volunteers
Expectancy bias
occurs when researcher knows which patients are distributed to which groups
researcher draws conclusions that support expected results
Measurement bias
method of collecting data influences results
Hawthorne effect
people act differently when they know they are being watched
Relative Risk
probability of outcome of interest in exposed group / probability of outcome of interest in non exposed group
