Biostatistics Flashcards
What is continuous data? Examples
Logical order with values that can increase or decrease by the same amount
Examples: Ratio and IntervalW
What is ratio data?
Absolute 0 where 0 means none
What is interval data
No meaningful zero
What is categorical data? Examples
Data fits into a limited number of categories
Ex: Nominal and ordinal
Nominal vs ordinal data
Nominal: Arbitrary order (order of categories doesn’t matter)
Ordinal: Categories can be ranked
What is the mean? When is it preferred?
Average
Continuous data with a normal distribution
What is median? When is it preferred?
Number in middle with data is organized numerically
Ordinal or continuous data that is skewed
What is mode? When is it used?
Most frequently data point
Nominal data
What is range?
The difference between highest and lowest values
What is standard distribution?
How spread out data is from the mean
What does a large standard dev mean?
Large amount of data is dispersed away from the mean
What is another name for bell-shaped curve?
Gaussian
Where do the central tendencies lie on a Gaussian distribution
Mean, median, mode are the same value
68% of data falls within 1 SD of the mean
95% of data falls with 2 SD
What makes a distribution skewed?
Number of values (sample size) is small or contains outliers
Method of measuring central tendencies in skewed data?
Median
What is the difference between outliers of high vs low values?
High: Right (positive) skew → mode to median to mean (left to right)
Low: Left (negative) skew → mean to median to mode (left to right)
What is the difference between independent and dependent variables?
Independent: can be manipulated by the research
Dependent: affected by the independent variables (outcomes)
What is the null hypothesis?
That there is no statistically significant difference between groups
Researchers aim to disprove or reject
What is an alternative hypothesis?
There is a statistically significant difference between groups
Researchers aim to prove or accept
What is the importance for alpha
Maximum permissible error margin
Commonly 5%
How do we interpret p-value using alpha (0.05)?
p-value < a: reject the null proving statistical significant results (alternative hypothesis accepted)
p-value > a: accept the null stating there is no statistical significance
What is a confidence interval
Provides information on the significance of p-value
CI = 1-a
When would you use CI of 0 vs 1?
Comparing differences of means: use 0
Comparing ratio data: use 1
How do you interpret CI when looking at means of data?
CI includes 0 → Not statistically significant
CI doesn’t include 0 → Statistically significance
How do you interpret CI when looking at ratio data?
CI includes 1 → Not statistically significant
CI doesn’t include 1 → Statistically significance
How do interpret a 0.95 CI 0.06, 9.35?
95% confidence that the true value lies between 6-35%
What is the difference between narrow and wide CI?
Narrow: high precision
Wide: poor precision
What is a false positive? How is it represented?
The alternative hypothesis was accepted when and the null was rejected in error
(Null is supposed to be accepted and no statistical significant difference present)
Represented as alpha (Type 1 error)
What is a false negative? How is it represented?
The null was accepted and the alternative hypothesis was rejected
(Alternative is supposed to be accepted → there is a statistically significant difference present)
Represented by beta (Type 2 error)
What is power?
The probability that a test would reject the null hypothesis correctly → the avoidance of type 2 errors
How do you interpret power?
More power → less likely of a type 2 error to occur
How do you calculate risk and risk ratio?
Risk = number of subjects of an unfavorable event/Total of subjects in group
Risk ratio = risk of treatment group / risk of control group
How do you interpret a risk ratio?
RR = 1 implies no difference in risk between the 2 groups
RR > 1 implies there is a greater risk in the treatment vs control
RR < 1 implies there is a greater risk in the control vs treatment
How to interpret a RR of .57?
The treatment group is 57% AS LIKELY to have the progression of outcomes as the control group
How do you calculate relative risk reduction?
RRR = (% risk of control - % risk of treatment) / % risk of control
RRR = 1-RR (as a decimal)
How do you interpret RRR of 47%
Treatment group is 47% less likely to have the outcome than the control
Why is absolute risk reduction more useful than RR or RRR?
RR and RRR is comparative data between treatment and control → no meaning of absolute risk
ARR includes reduction of risk and the incidence rate of outcomes
How do interpret a study if ARR was not reported?
Risk reduction is minimal to start with, therefore ARR would have been small
How do you calculate ARR?
I % of control - % of treatment I
How do you calculate NNT and NNH?
1/ (risk in control - risk in treatment)
1/ARR (in decimal)
NNT: Round up
NNH: Round down
How do interpret a NNT of 9 in a study that lasted for 1 yr
9 patients needed to receive treatment for 1 yr to see a progression prevented in 1 patient
What is NNT?
Number of patients needed to treat for a period in order to see a benefit in 1 patient
What is NNH?
Number of patients needed to be treated for a period in order to see one patient harmed
How do interpret a NNH of 9 in a study that lasted for 1 yr
9 patients needed to receive treatment for 1 yr for 1 patient to be harmed
What study primarily uses odds ratio?
Case control
What are odd ratio?
The estimated risk of an unfavorable event associated with a treatment of intervention
How do you calculate OR?
OR = (# outcome with exposure/ # no outcome with exposure) / (# outcome without exposure / # no outcome without exposure)
What data is used in survival analysis?
Hazard rate
How do you calculate hazard ratio
Hazard rate in treatment / Hazard rate in control
How do you interpret OR and HR?
OR or HR = 1: the event rate is same (no advantages to the treatment)
OR or HR >1: the event in the treatment group is higher than the control group (control is more advantageous)
OR or HR <1: the event is the control group is higher than treatment (treatment is more advantageous)
What is the difference between primary and composite endpoints?
Primary: main result that is measured for significant benefit
Composite: combines individual endpoints into one measurement
What is the criteria of making a composite endpoint?
All endpoints must be of similar magnitude and similar importance to the patient
What is the difference between parametric and nonparametric test?
Parametric requires normal distribution
Nonparametric: for data not normally distributed
Differentiate the types of t-tests? What are they for?
T-tests are used for continuous data that is normally distributed
One-sample: 1 sample
Dependent/paired: 1 sample (before and after)
Independent/unpaired student: 2 samples (treatment and control)
What test is used for continuous data with a normal distribution and ≥3 samples?
Analysis of variance: ANOVA (F-test)
What is a sign test
Skewed distribution of 1 sample
What is Wilcoxon signed rank test for?
Continuous skewed distribution of 1 sample with 2 measurements
Categorical data of 1 sample with 2 measurements
What is Mann-Whitney test used for?
Continuous Skewed distribution of 2 samples
Ordinal data with 2 samples
What is Kruskal-Wallis test?
Continuous Skewed distribution of ≥3 samples
Categorical data of ≥3 samples
When is chi-square tests used?
Categorical data of 1 or 2 samples
When is Fisher’s extract used?
Categorical data of 2 samples
