Statistics Flashcards
How do you calculate Absolute Risk Reduction (ARR)
ARR = Control rate - experimental rate
(If a control has a risk of PONV 20/100 compared to an experimental drug 10/100 the ARR = 0.2 - 0.1)
How do you calculate Number Needed To Treat (NNT)
1/ARR
(Absolute risk reduction)
How do you calculate Relative Risk (RR)
RR = Experimental event rate / Control event rate
(If a control has risk of PONV 20/100 and experimental drug has risk 10/100 RR= 0.1/0.2)
What type of error is a false positive
Type 1 error (alpha)
What type of error is a false negative
Type 2 error (Beta)
How do you calculate power of a study
Power = 1 - Beta (Type 2 error)
Which statistical test can be used for comparing 2 sets of continuous normally distributed (parametric) data
T-test
Paired if the same subject is in the control group and also in the experimental group
otherwise unpaired
Which statistical test can be used for comparing more than 2 sets of continuous normally distributed (parametric) data
ANOVA test
(Can do multiple t-tests but higher risks of error)
What percentage of a population lies within 1 standard deviation (SD) from the mean
68%
What percentage of a population lies within 2 standard deviations (SD) from the mean
95%
What percentage of a population lies within 3 standard deviations (SD) from the mean
99.7%
How does the size of the difference you’re looking for between 2 groups affect power
Small difference = low power
High Beta error ( false negative)
Needs high n number
How does the SD affect power
Large SD = large area = low power
high Beta error
Needs high n Number
How does the p value affect power
low p value = high confidence can reject null hypothesis = low power
High Beta error
Needs high n number
How does parametric vs non-parametric test affect power
Non-parametric = low power
high beta error
Needs high n number
How do you calculate standard error of the mean
SD/ √n
standard deviation/ square root of sample size
How does standard error of the mean give you 95% CI
95% confidence interval is 2 standard errors of the mean above and below.
How do you calculate variance
Σ(x-x̄)²/ n-1
the difference of each value from the mean added together and squared then divided by the sample size -1 (degree of freedom)
How do you calculate standard deviation from variance
SD = √variance
= √Σ(x-x̄)²/ n-1
What does it mean if the 95% confidence intervals of 2 groups do not overlap
There is likely to be a significant statistical difference
Which statistical test can be used for comparing more than 2 groups of quantitative non-parametric data
Kruskal wallis test
Which 2 statistical tests can be used for comparing 2 groups of quantitative non-parametric data
Mann Whitney
Wilcoxian
Which parameter is used to look at the central tendency of non-parametric data
Median
(use median instead of mean)
Which parameter is used to look at the central tendency of parametric data
Mean
Which parameters are used to look at spread of non-parametric data
Range
Inter quartile range (from start of Q2 (25%) to end of Q3 (75%) )
Which 2 statistical tests can be used for comparing 2 groups of qualitative data
Fisher’s exact test
Chi- Squared test
What does a correlation coefficient of r = +1 mean
perfect positive linear correlation. All data points lie exactly on a straight line with +ve gradient
What does a correlation coefficient of r = -1 mean
perfect negative linear correlation. All data points lie exactly on a straight line with -ve gradient
What does a correlation coefficient of r = 0 mean
No linear correlation between data points
When would you use Pearson’s correlation coefficient
When comparing 2 groups of parametric data to look for degree of correlation with linear relationship
When would you use Spearmen’s rho correlation coefficient
When comparing 2 groups of which at least 1 has non- parametric data to look for degree of correlation with linear relationship
Define what is meant by co-efficient of determination (r²)
(This is correlation co-efficient squared)
The proportion of the variation (variance) in the dependent variable that is predictable from the independent variable
e.g r ² of 0.64 means that 64% of the variance in y is predictable from the variance in x
What is the Bland-Altman plot used for
Used to compare 2 different measuring devices
X-axis: Both measurement device results added together and divided by 2 (average) (the true value)
Y-axis: Difference between the values measured by each device
How do you calculate Relative Risk Reduction (RRR)
RRR = (Control event rate - Experimental event rate) / Control event rate
(If a control has risk of PONV 20/100 and experimental drug has risk 10/100
RRR= (0.2 - 0.1)/0.2
When would you use odds ratio (OR) instead or relative risk (RR)
Case control studies
RR can only be used in cohort studies whereas OR can also be used in cohort and case control
What does an odds ratio (OR) of 1 mean
There is no association between exposure to the treatment and improved outcomes
Null hypothesis
What does an odds ratio (OR) < 1 mean
There is a negative association
The odds of an event happening in the exposed/treated group is less than the control group
(e.g treatment reduces odds of stroke)
What does an odds ratio (OR) > 1 mean
There is a positive association
The odds of an event happening in the exposed/treated group is higher than the control group
(e.g treatment increases odds of stroke)
