Biostatistics Study Cheat Sheet

Summary of Data Types

Continuous
Nominal Ordinal
Interval Ratio
Data ranked in a specific Data ranked in a specific
Data without Data ranked in a
order that includes a order that includes a
order or specific order but
Definition: constant difference in constant difference in
indication of without a consistent
magnitude of change. Zero magnitude of change. Zero
relative severity magnitude
is arbitrary is NOT arbitrary.
Examples: Sex, mortality Trauma score Temperature Blood pressure, heart rate
Eligible for
No No If normally distributed
parametric tests:
Typical measure
of central Not applicable Median Mean / average
tendency:

EQUATIONS
Outcome Yes Outcome No
Intervention Group A Group B
Control Group C Group D

Intervention event rate = A/B Control event rate = C/D

OR = (A/C) / (B/D) or (AD/BC) RR = A/(A+B) – C/(C+D) RRR = 1 – RR

ARR = A/(A+B) – C/(C+D) NNT = 1/ARR Power = 1 – β

If the... …it will… When the... …it means…

Make the study require less Data being produced by a
Sample size decreases
resources to perform project is intended to
It is a research project
Sample size increases Increased the power contribute to generalizable
Power increases Make results more reliable knowledge
Confidence interval Study poses any threat to Institutional review board
Reduce uncertainty human subjects approval is required
narrows
Reduce likelihood outcomes Conclusion is that there is a
P-value gets smaller A type I error has
occurred by chance difference when there actually
occurred
Standard deviation gets Imply reduced variability in is no difference
smaller the sample Conclusion is that there is no
A type II error has
Data is not normally Not allow for parametric difference when there actually
occurred
distributed tests to be used is a difference
Make the outcome less likely Subject remembers an event
Odds ratio is < 1 There is recall bias
in the intervention group differently from how it occurred
Suggest a negative Study subjects alter their
The Hawthorne Effect is
Relative risk is < 1 association between the behavior due to awareness of
occurring
intervention & outcome being in a study
Mean the intervention is less Study design is not consistent The study has limited
ARR becomes less with standard clinical practices external validity
meaningful
Increase the chance of a Study is observational in There is no intervention
Alpha increases design group
type I error
Increase the chance of a Study is a snapshot of a It is a cross-sectional
Beta increases population at one point in time study
type II error
Not impact the median as Design is a case-control study It is a retrospective study
Study has extreme Use of multivariate logistical Independent variables
much as it would impact the
outliers regression is employed can be identified
mean

Abbreviations: ARR = absolute risk reduction, NNT = number needed to treat, OR = odds ratio, RR = relative risk,
RRR = relative risk ratio

[Last updated June 2018] For More Resources Like This Check Out: www.training.IDstewardship.com