Statistics Flashcards
What is, and what are the types of qualitative (categorical) data?
Qualitative - each individual can only belong to one of a number of distinct categories.
Binary of two categories - male/female
Nominal: categories with names but without order
- O, A, AB, B blood groups
Ordinal data: an order exists to categorise
- Cancer staging, pain score, ASA score
What is, and what the types of quantitative (numerical) data?
Variable has a numerical value
Parametric data: continuous numerical data from a normally distributed population
Non-parametric data: non-normal distribution, or when sample size is too small
Interval data (not true zero) and ratio data (true zero) used to describe temperature
What is mean?
average of the sum of observations
What is median
Middle of the series of observations
What is the mode?
Value that occurs most frequently.
Define type I error
False positive - frequency where we erroneously conclude there is a difference when there isn’t one
Determined by the alpha value, usually set at 5%
Define type 2 error
False negative - frequency where we are unable to detect a difference when there is one
Define p value
Probability of finding this result by chance if the null hypothesis is true.
Probability of this being a false positive result
Define sensitivity
Probability that a positive result indicates the presence of finding.
I.e: high Mallampati score = difficult airway
Sens = true pos / (true pos + false neg)
high sens = low false neg rate
Define specificity
Probability that a negative result indicates the absence of the finding
Spec = true neg / (true neg + false pos)
High spec = low false pos rate
Define positive predictive value
Probability of a positive finding when the test is positive
PPV = true pos / (true pos + false post)
Define negative predictive value
Probability of a negative finding when the test is negative
NPV = true neg / (true neg + false neg)
How would low incidence rate affect
- Sens
- Spec
- PPV
- NPV
Low incidence of, for example, difficult airway of 1/2000, with regard to MP testing
- No effect on sensitivity and specificity as they are inherent properties of the test
- Low PPV due to low true pos
- High NPV
Limitations of P value?
Selection of 0.05 is totally arbitrary and has no clinical basis
Statistical significance does not equal clinical significance
Presentation of p value of <0.05, rather than exact value, prevents the reader from interpreting the degree of significance.
What is confidence interval?
a range of sample data which contains an unknown population parameter, such as the median or mean.
If 95% interval is used, this implied that if the study is repeated numerous times, the quoted range will contain the unknown population parameter 95% of the time
Define relative risk
the risk of the event in the intervention group compared with the risk of the event in the control group
Amplifies the apparent effect of a drug on rare outcomes
- Risk of 1% vs. 0.3%
- Absolute risk reduction = 0.7%
- Relative risk = 70%
Define odds ratio
A ratio of event to non-event in the intervention group compared with the control group
Define Hazard ratio
The relative risk of an event happening at time t
- Risk of pain now compared to risk of pain at some point
How to calculate number needed to treat?
NNT = 1/ Absolute risk reduction
Which statistical test would you use for normal data, whether it be paired or unpaired?
Sample T-test
When would you use the Mann-Whitney U test?
For unpaired, non-parametric data
i.e height of male vs. female, when the sample size is small
When would you use the Wilcoxon Matched Pairs test?
Paired, non-parametric data
i.e study of a pain medication on small sample size of 10 patients.
Each patient provides and pre and post intervention pain score
What tests would you use for study of more than 2 groups?
If normal distributed (parametric) - use ANOVA
If not normal
- Paired data = Friedman test
- Unpaired data = Kruskal-Wallis Test
What tests could you use for categorical data?
Chi Squared test compares the distribution of a categorical variable between two or more independent groups.
- Versatile, used for larger sample sizes. Doesn’t calculate exact P value
Fisher’s exact test
- For smaller sizes, or when expected frequencies are low
- For unpaired data
- Can calculate p value
McNemar’s test
- For paired data, calculates exact p value
