Statistics Flashcards
Validity
extent to which something measures what it claims to measure.
Internal Validity
Confidence that we can place in the cause and effect relationship in a study.
Ie something else did not cause the change
Reliability
extent to which an experiment, test, or any measuring procedure yields the same result on repeated trials.
External validity
The degree to which the conclusions in a study would hold for other persons in other places and at other times, i.e. its ability to generalise.
Standard Deviations
How spread out is the data from the mean
1: 68.2
2: 95.8
3: 99.7
How can you read skewed data?
IQ ranges
Median
Box and whisker plots
Used for skewed data
IQ ranges and median
Box and whisker plots
Mann Whitney U test
Non parametric
compares ordinal, interval, or ratio scales of unpaired data
Wilcoxon test
Before and after test on same population
Student t test
normally distributed
paired data – from the same group of patients
unpaired - different groups
Pearsons, Spearmans
Pearsons - norm dis
Spearmans - not
Correalation
Chi squared
non parametric
between 2 groups
compares percentages/ proportions
Type 1 error
the null hypothesis is rejected when it is true
T2 Error
the acceptance of the null hypothesis when it is actually false
Power of a study
the probability of (correctly) rejecting the null hypothesis when it is false, i.e. the probability of detecting a statistically significant difference
power = 1 - the probability of a type II error
Hawthorne Effect
People change their behaviour because they know theyre being studied
Parametric data
Normally distributed
Parametric test examples
Students T test
Pearsons
Likelihood ratio (positive)
Sens/ 1- spec
Likelihood ratio (negative)
1-sens/ spec
Absolute risk
CER - EER
Relative risk
EER/ CER
Relative risk reduction
Absolute risk/ CER
Number needed to treat
1/AR
Pearson test
Normally distributed
Correlation
Funnel plot
At the top is the more powerful studies
Shows bias
Odds ratio
Odds of experiment/ odds of control
Power
1- T2 error
P value
probability of getting a result by chance in the null hypothesis is true
Anova test
statistically significant difference between the means of several groups
normal distribution
compares variance of 2 means
similar to t test
Odds ratio
Odds of signicant event in variable/ odds of signicant event in control
Cohort study
Measures relative risk normally
Observational and prospective
2+ selected due to exposure to a particular agent
And followed up to see disease outcome
Case control study
Observational and retrospective
Patients with a condition are identified and
matched to controls
Data is collected on past exposure to find cause
Outcome measured in odds ratio
Cheap
Prone to confounding
Standard error of the meann
standard deviation/ square root of number of patients
ie increases if smaller populations
confidence interval
95% chance mean is in this
Lower limit of the mean
1.96 x SEM
Purposive sampling
Preselected criteria
Relevant to research questions
Quota sampling
type of purposive sampling
Decide while designing study the number and characteristics
Snowball
Chain referral - patients tell other patients
Reach hidden populations
Convenience sampling (opportunistic)
Open period of recruitment - first come first served, up until a set number
Hawthorne bias
People change behaviour as they known theyre being studied
