Biostatistics Flashcards
Types of Data: Can be _________ or ________
continuous or discrete/caterogorical
Continuous data can be _________ or ________
Discrete/categoircal data can be _______ or _______
Continuous: ratio or interval
Discrete: nominal or ordinal
what are measures of central tendency?
mean, median, mode
mean is best for what kinds of data?
continuous or normally distributed values (no outliers!)
median is best for what kind of data?
ordinal data or if there is SKEWED continuous data
mode is best for what kind of data?
nominal data
what is gaussian?
it is normal bell shaped distribution curve
Gaussian distributions:
____% of values fall in 1 SD
____% of values fall in 2 SD
1 SD: 68%
2 SD: 95%
alpha level:
used to select the maximum permissible ________
error margin
alpha level:
commonly set at _____
5% of 0.05
if they do 1% of 0.01 — just need more data to prove this
if the p-value is less than alpha: the null hypothesis is _______ and the result is termed statistically _______
null is rejected
statistically significant
Confidence Interval = _____ - _____
1 - alpha
alpha of 0.05 = CI of 95%
The CI range will not include _____ when comparing difference data/means
0
The CI range will not include _____ when comparing ratio data (hazard ratio, odds ratio, hazard ratio)
1
Type 1 errors: False _______
Type 2 errors: False ______
1: false POSTIVES – we rejected the null when we shouldn’t have
2: false NEGATIVES – we accepted the null when we should have rejected it
Type 1 errors are related to (alpha or beta)
Type 2 errors are related to (alpha or beta)
1: alpha
2: beta (type 2 error = beta)
Study power: 1 - ____
Power = the probability we will _____ the null hypothesis correctly
Beta (type 2 error)
we will REJECT correctly (aka power to avoid type 2 error)
Risk – how to calculate?
vs
Relative risk — calculate it how?
Risk: # of subjects in group to have an event / total # of subjects in group….
RR: risk in exposed group (treatment)
divided by
risk in the control group
Relative Risk Interpretation
RR = 1
RR > 1
RR < 1
1: no difference in outcome b/w groups
RR > 1 = greater risk for outcome in tx group
RR < 1 = lower risk/reduced risk of outcome in tx group
Relative risk vs relative risk reduction?
RR: determines where the risk is less or more
RRR: indicates how much the risk was reduced
How to find RRR? (relative risk reduction)
1 - RR
OR
% in control - % in tx group / % in control group
ARR what is it and how to calculate it?
it is the reduction in risk and the incidence rate of the outcome
ARR = % risk in control group - % risk in tx group
NNT what does it mean and how to find it?
NNT = number need to treat = how many patients need to be treated in a certain time frame for ONE patient to benefit
NNT = 1 / ARR
or NNT = 1 / (risk in control group - risk in tx group)
For NNT calculations – if you get a fraction – round up or down?
always round up (this is conservative)
For NNH calculations – if you get a fraction – round up or down?
always round down (this is conservative)
NNH what does it mean and how to find it?
Number needed to harm = how many patients need to be treated in a certain time frame for ONE patient to be HARMED
NNH also is 1 /ARR (same as NNT)
_______ test is good for nominal or ordinal data
chi square
sensitivity the true _______
vs
specificity the true ______
sensitivity: true positive
specificity: true negative
intention to treat vs per protocol?
intention: all pts originally put in the tx group are still included even when they did not complete it
per protocol: all include pts that completed the study
