Browse the glossary using this index

Special | A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | ALL

Page:  1  2  3  4  5  (Next)
  ALL

A

Alpha-Level (α-level)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Significance level
  • Probability of rejecting H0 when H0 is true
  • Usually set at α=0.05 or α=0.01

Alternative Hypothesis (H1 or Ha)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  •  Experimental hypothesis
  • The population parameter differs from H0
  • When H0 can be rejected, it does not mean that H1 can be accepted, because of the remaining probability of α- and β-error

 


B

Bias

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Systematic error or deviation in results
  • Example: selection bias

Systematic differences in the groups that are compared: group A has 50% more smokers than group B 

→ groups are not comparable



Blinding

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Process of hiding which comparison group a particular participant belongs to
  • Used to minimize risk of bias
  • Common methodology in clinical trials
  • Types:
  1. Single blind: participants unaware
  2. Double blind: participants+ outcome assessors/ testers unaware
  3. Triple blind: participants+ outcome assessors/ testers+ data analysts unaware

 


C

Confidence Interval (CI)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Measure of uncertainty around main finding of statistical analysis
  • Usually set at CI=95%, also CI=99% or CI=90%

 


Continuous Data

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Information that can take any value within a certain range
  • Opposite of discrete data
  • Example: Weight of a person: The weight can be any value within the range of people´s weight (e.g. 60.3 kilogram).


D

Dependent or Outcome Variable

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

Clinical trial

Dependent variable= Outcome (ill or healthy)

Independent variable= Treatment arm (new drug)

 


Discrete Data

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Information that can only take certain values
  • Opposite of continuous data
  • Example: Number of people in a room: It is impossible to have 40.3 people in a room.


F

False Negative

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

= Negative test when individual actually has the disease



False Positive

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

= Positive test when individual does not actually have the disease



H

Hypothesis

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

= statement being tested


I

Incidence

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Number of new occurrences of a condition in a population over a given period of time
  • Example: number of new cases of breast cancer in Bavaria over one year


Independent Variable

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

Clinical trial

Dependent variable = outcome (ill or healthy)

Independent variable = treatment arm (new drug)



Interval Scale

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Measurement scale; interval known, but no true “0” point
  • Type of data: metric, quantitative, discrete or continuous
  • Example: temperature (Fahrenheit or Celsius), IQ, etc.

 


L

Levels of Evidence

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

Level I:           Evidence from a systematic review or meta-analysis of all relevant RCTs

Level II:          Evidence obtained from well-designed RCTs

Level III:         Evidence obtained from well-designed controlled trials without randomization

Level IV:         Evidence from well-designed case-control and cohort studies

Level V:          Evidence from systematic reviews of descriptive and qualitative studies

Level VI:         Evidence from single descriptive or qualitative studies

Level VII:        Evidence from an opinion of authorities and/ or reports of expert committees

Further information: Evidence Based Practice Toolkit

More information about different study types: Literature Reviews: Types of Clinical Study Designs 


M

Maximum (MAX)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Largest number in a data set
  • Example:
Set of numbers: 3,4,5,5,5,6,7

MAX= 7


Mean (μ, x̄)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • (Arithmetic) average
  • Sum of all observations divided by the number of observations
  • Formula:   \mu = (x1+ x2+ …+ xi)/ n
  • Example: 

Set of numbers:  \mu  = 3, 4, 5, 5, 5, 6, 7 = (3 +4 +5 +5 +5 +6 +7)/ 7 = 35/ 7= 5

 

 


Median

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Value of the observation that falls in the middle when observations are ranked in order
  • Formula: 
  1. Uneven number of observations: Median= x(n+1)/2
  2. Even number of observations: Median= (x(n/2)+ x(n+2)/2)/ 2
  • Example: 

Order all observations from smallest to highest number

  1. Uneven number of observations: n= 7 
Set of numbers (ordered from smallest to biggest number): 3, 4, 5, 5, 5, 6, 7 
Median: x(n+1)/2 = x(7+1)/2= x4→ x4= 5
 
2. Even number of observations: n= 8

Set of numbers (ordered from lowest to highest number): 3, 4, 5, 5, 6, 6, 7, 9

Median: (x(n/2)+ x(n+2)/2)/ 2= (x(8/2)+ x(8+2)/2)/ 2= (x4+ x5)/ 2= (5+ 6)/ 2= 11/ 2= 5.5

 

 

 


Minimum (MIN)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Smallest value in a data set
  • Example:
Set of numbers: 3,4,5,5,5,6,7
MIN= 3




Mode

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Most frequent value in a data set
  • Example:
Set of numbers: 3, 4, 5, 5, 5, 6, 7

Mode= 5


N

Nominal Scale

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Classification scale; each person is assigned to one “type”
  • Type of data: categorical, qualitative, discrete
  • Example: blood type (A, B, 0), type of breath sound, type of arthritis, etc.

 


Normal Curve

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

Normal Curve

 


Normal Distribution

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Data that is symmetrically distributed around a mean value
  • Description of distribution via mean and standard deviation

Null Hypothesis (H0 or Ho)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Statistical hypothesis
  • The population parameter is equal to the claimed value

Number Needed to Treat (NNT)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • The number of patients that need to be treated before one person would experience the desired outcome
  • Number needed to treat to benefit: number of patients that need to be treated before one person would experience a beneficial outcome
  • Example: NNT= 20: this means that 20 people need to be given a stroke prevention drug before one stroke can be prevented
  • Number needed to treat to harm: number of patients that need to be treated before one person would experience a harmful outcome
  • Example: NNT= 40: this means that 40 people can be treated with cardiac surgery before one person dies 

O

Objectivity

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Data collection method which will always come to the same result
  • Example:

Interviewer A asks question to Patient X

Interviewer B asks question to Patient X

→ Both answers will be the same (answer should be independent from the interviewer)


Odds Ratio (OR)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Chance of an event occurring; this is calculated by taking the number of individuals in the sample who experience the event divided by the number of individuals for whom the event did not occur
  • Mainly used in case-control studies
  • Small risk → Odds Ratio ~ Relative Risk
  • Ranges:
    1. OR> 1.0 – exposure may increase the odds of a disease
    2. OR> 1.0 – exposure does not affect the odds of outcome
    3. OR< 1.0 – exposure may decrease the odds of a disease
  • Example: OR= 3.7 for likelihood to die from a new antihypertensive drug as compared to an existing drug

→ Patients who received the new antihypertensive drug die 3.7 times more often than patients who received an existing drug.

→ The odds to die with the new antihypertensive drug is 3.7 times higher than with the existing drug.

Further information: Explaining Odds Ratios


Ordinal Scale

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Ranking scale; no known interval
  • Type of data: categorical, semi-quantitative (= quantitative but open to individual interpretation), discrete
  • Example: pain levels, joint laxity grades, Manual Muscle Testing grades, level of assistance

 


P

Percentile

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Value that a certain percentage of data falls below
  • Example: 

50th percentile= 50% of all values in a distribution fall below this score

 


PICO

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

P=       Population

I=        Intervention (or diagnosis, prognosis) being evaluated

C=       Comparison (usually to gold standard or no treatment)

O=       Outcome

Example:

Is physical activity at least twice a week for more than ten minutes as effective as the antihypertensive X in preventing high blood pressure in adults over 18 years?

P=       adults over 18 years

I=        physical activity at least twice a week for more than ten minutes

C=       antihypertensive X

O=       prevention of high blood pressure


Prevalence

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Proportion of a population having a particular condition at a specific point of time
  • Used to look at burden of disease in a population
  • Example: percentage of men over 40 with lung cancer today in Bavaria

P-Value

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Probability that the statistical results occurred by chance
  • When p-values is smaller than the stated level of significance (α-value), H0 is rejected 

Q

Qualitative Data

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Number of subjects: small
  • Type of data collection: observations, patient interviews, verbal interactions
  • Objective: to generate hypothesis, exploratory and not conclusive
  • Example: Physicians’ opinion about vaccination

Quantitative Data

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Number of subjects: large
  • Type of data collection: questionnaires, counting, experimental trials
  • Objective: to test a hypothesis, used to recommend course of action
  • Example: Mean time youths spend with their mobile devices a day

Quartile

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Distribution divided into four equal parts
  • Q1= 25th percentile (score at which 25% of the distribution’s value fall below)
  • Q2= 50th percentile
  • Q3= 75th percentile


R

Randomized-Controlled Trial

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Experiment which compares two or more interventions 
  • Intervention is compared to control intervention (= no intervention or another type of intervention)
  • Participants for each intervention are randomly assigned
  • Common assignment to group: one intervention assigned to each individual
  • Cluster-randomization: assignment to defined groups of individuals (e. g. class, household)
  • Example:

Randomized-Controlled Trial: Treatment success for patient A with drug A in comparison to treatment success for patient B with already existing drug B

Cluster-Randomised-Controlled Trial: Prevalence of obesity in a community with a systematic physical activity program versus a community without any such program


Range (R)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Difference between lowest and highest value
  • Formula: R= Maximum- Minimum
  • Example:

Set of numbers: 3,4,5,5,5,6,7

R= 7- 3= 4







Ratio Scale

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Measurement scale; interval known with 0 point
  • Type of data: metric, quantitative, discrete or continuous
  • Example: BMI, cigarettes per day, age, etc.

 


Relative Risk (RR)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Risk of a certain event happening in one group versus the same event happening in another group
  • Usually related to risk factor exposure group versus control group
  • Ranges:

 

 

 

 

  1. RR> 1.0 – exposure increases risk
  2. RR= 1.0 – equal probability in both groups
  3. RR< 1.0 – exposure decreases risk

 

 

 

 

 

 

 

 

  • Example: RR= 1.3 for lung cancer in smokers compared to non-smokers
→ Smokers have a 1.3 higher risk of getting lung cancer than non-smokers

 

 

 


Reliability

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Reproducibility of a study
  • Example:

First round of an experiment 

Second round of the experiment under the same conditions

→ Both experiments have the exact same outcome


Risk Factor

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Aspect of a subject’s condition, lifestyle or environment that affects the probability of a disease to occur 
  • Example: obesity = risk factor for high blood pressure

S

Sensitivity

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Measure of a test’s ability to correctly detect people with the disease
  • Sensitivity and specificity are inversely proportional: when sensitivity increases, specificity decreases and vice versa
  • Example: 

A sensitivity of 97.5% in mammography screening means that every woman with a tumor will be correctly detected in 97.5% of the cases. 2.5% might have a tumor, but cannot be identified through the screening.

Further information:  Understanding and using sensitivity, specificity and predictive values.

Example for using sensitivity in a study:  Sensitivity and specificity of mammography and adjunctive ultrasonography to screen for breast cancer in the Japan Strategic Anti-cancer Randomized Trial (J-START): a randomised controlled trial. 


Skewed Distribution

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

  • Asymmetrical shape of distribution
  • Types
  1. Skewed to the left: negative skew (long tail on the negative side of the tail)Skewed left
  2. Skewed to the right: positive skew (long tail on the positive side of the tail)skewed right



Specificity

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Measure of a test’s ability to correctly identify people who do not have the disease
  • Sensitivity and specificity are inversely proportional: when sensitivity increases, specificity decreases and vice versa
  • Example: 

A specificity of 95.0% in mammography screening means that 95.0% of the women tested are correctly identified as not having a tumor.

Further information:  Understanding and using sensitivity, specificity and predictive values


Standard Deviation

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Measure of the spread/dispersion of a set of observations
  • Square root of the variance
  • Formula:   s= \sqrt{(\sum\nolimits_{i=1}^n(x_i- \mu)^2)/(n-1) }
  • Example:

Set of numbers: 3,4,5,5,5,6,7

 



Statistically Significant

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Result that is unlikely to have happened by chance
  • It is assessed by the p-value

T

Type-I-Error (α-error)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 

  • Incorrectly rejecting the null hypothesis
  • Concluding that there is a relationship when no relationship exists
  • Example: if α=0.01, then there is a 1% chance for an α-error
  • False Positive

Type-II-Error (β-error)

(Last edited: Tuesday, 14 March 2017, 8:55 AM)

 

 


V

Validity

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Degree to which a result is likely to be true and free of bias
  • Test is measuring what is needed
  • Accuracy of a test
  • Example:

An interview guideline for breast cancer experts contains questions about the topic "breast cancer". This guideline includes previously defined research questions regarding the topic so that it will help in answering the required questions. 

Further information: Understanding and using sensitivity, specificity and predictive values 


Variance

(Last edited: Tuesday, 14 March 2017, 8:55 AM)
  • Measure of variation shown by a set of observations
  • Sum of the squares of deviations from the mean divided by the number of observations minus 1 (because it is a sample)
  •  s^2=( \sum\nolimits_{i=1}^n (x_i- \mu)^2)/(n-1)
  • Example:

Set of numbers: 3,4,5,5,5,6,7





Page:  1  2  3  4  5  (Next)
  ALL