Basic statistics Flashcards
What are the 4 scales data can be described?
Nominal - naming scale
Ordinal - ordering scale 1st, 2nd , 3rd
Interval - Scale 1,2,3,4 or 5,10,15,20
Ratio - interval scale with a fixed 0
The shapiro-wilk test and the kolmogorov-smirov test are tools to assess what?
Assessing normality of data ie if normally distributed
How to calculate median
(n+1)/2
How to calculate mean
Total sum/n
Is median or mean more influenced by skewed data?
Mean
How do you quantify variability in a normal distribution?
Variance
Standard deviation
Standard error of mean
Confidence interval
What is variance
The average amount by which any individual measurement differs from the mean
What is standard deviation, why is it used?
Square root of the variance (in ordinal units)
As some differences will be positive and negative, differences cancel others out which is logically useless -
What is the standard error of the mean? What does it take into consideration?
How close a sample mean is to the population mean
SEM: Standard deviation/sqaure root of sameple size
What transformations of data can be used it the data is not noramally distrubuted?
Logarithm of the data
Square root
Squaring
reciprocal 1/x
Binomial distribution is a scale for what kind of data?
Discontinous dichotomas e.g. yes/no, dead/alive
If proportion of one category is know the pther can be calcualted 1-p
In Binomial distribution when number of observations is needed for a normal distibution
n>30
Poisson disctribution is a distribution for which kind of data
dicrete data. arising when the number of occurances of an event per unit of time are counted
e.g. number of admission per day
It is skewed if the mean is close to 1
The vary of propability is between what
0-1
What are type one errors
When the hypothesis is wrongly rejected (if p<0.05, once in 20 trials)
what are type two errors
When the hypothesis is wrongly not rejected
What is power
Power is the probability of rejecting the null hypothesis when, in fact, it is false. (Avoiding type 2 error)
Dependant on sample size
Which analysis can be used for testing associations between groups in non-parametric data?
Spearmans rho
Which analysis can be used for testing associations between groups in normally distributed data?
Pearsons
Linear regression analysis
Multiple linear regression analysis
Which analysis can be used for comparing groups in non-parametric data?
Sign test Chi square MnNemar test Wilcoxon test Mann Whitney U test Kruskal-Wallit analysis
Which analysis can be used for comparing groups in normally disributed data?
Student T test, paired and unparied
F test
Analysis of variance
When can Chi squared not be used?
Proportion
Any expected number of observations < 5
total number of observations < 20
What is the T-test used for?
Compare
- the mean of a sample with a predited value
- the means of two sample
Must be normal distribution
In the T-test what does it mean if the confidence interval crosses 0.
There is no significant difference between the means of the 2 groups at the specified level of confidence
What is persons correlations coeficient used for>
Measure the degree of assocaiation between 2 variable and is a measure of the scatter of the data around an underlying trend.
At least 1 set of data must be normally distrubuted
In persons correlation coefficient, what does the values of 1, -1 and 0 mean?
1 - perfect positive correlation
-1 - perfection negative correlation
0 - no association
What is sensitivity?
Proportion of true positives that are correctly identified
What is specificty?
Proportion of true negatives that are correctly identifed
Positive predictive value?
proportion of people with positive results who are correctly identified
Negative predicitive calue
proportion of people with a negative result who are correctly identifed
What is incidence?
The rate at which new cases of the disease occur in a population previously free of it
What is prevalance?
The frequency of existing disease at a given time
What is an odds ratio?
OR gives an estimate for which a confidence interval can be calculated for the association between 2 binary variables and allows for testing of other variables on thar assoication (logisitic regresstion)
What is relative risk?
The relative risk (RR) or risk ratio is the ratio of the probability of an outcome in an exposed group to the probability of an outcome in an unexposed group.
Must be selected prospectively.
What is level 1a evidence?
Systematic review and meta-analysis of RCT
What is level 1b evidence?
At least 1 RCT
What is level 2a evidence?
At least 1 well designed control study, without randomisation
What is level 2b evidence?
At least 1 other type of well designed quasi-expereimental study
What is level 3 evidence
Well designed descriptive studies
What is level 4 evidence
Expert committee reports/opinions and/or clinical experience of respected authorities.
