Statistics/Evidence Flashcards
Levels of evidence for intervention studies?
1. 1++
2. 1+
3. 1-
4. 2++
5. 2+
6. 2-
7. 3
8. 4
- 1++
- High quality meta analyses, systematic reviews of RCT’s or RCT’s with very low risk of bias - 1+
- Well conducted meta- analyses, systematic reviews of RCTs, or RCTs with a low risk of bias - 1-
- Meta analyses, systematic reviews of RCTs, or RCTs with a high risk of bias - 2++
- High quality systematic reviews of case control or cohort studies; high quality case-control or cohort studies with a very low risk of confounding bias or chance and a high probability that the relationship is casual - 2+
- Well conducted case control or cohort studies with a low risk of confounding, bias or chance - 2-
- Case control or cohort studies with a high risk of confounding bias - 3
Non analytical studies (eg. Case reports, case series) - 4
Expert opinion, formal consensus
Z Values
1. Define
Important component of confidence intervals.
Measures the number of standard errors to be added and subtracted in order to achieve your desired confidence level.
80%z = 1.28
85%z = 1.44
90%z = 1.64
95%z = 1.96
98%z = 2.33
99%z = 2.58
Parametric data vs non parametric data
Parametric test eg. weight/age
- normal distribution of data
- data continuous
- independence of data (one group does not influence another)
- homogeneity
- considered more powerful tests
Non-Parametric Test:
- no assumptions with regards to distribution
- data can be continuous or ordinal
- can be transformed to parametric data
Examples of parametric tests?
- Paired t test
- Unpaired t test
- Pearsons
- Multiple regression
Examples of non parametric tests?
- Mann-Whitney U test
- Wilcoxon matched pairs test
Standard Deviation
Represents spread of the population
Odds Ratio
Represents the odds that a diseased group were exposed, compared to the odds of an undiseased group (controls) being exposed.
Eg. Smoking and lung cancer
Odds ratio
OR = 1
OR > 1
OR < 1
OR = 1 - no difference in the odds of exposure between the two groups
OR > 1 - diseased group more likely to have been exposed compared to controls
OR < 1 - diseased group less likely to have been exposed compared to controls
Calculating odds ratio
2x2 table
A = exposed patients, outcome +ve
B = exposed patients, outcome -ve
C = unexposed patient, outcome +ve
D = unexposed patient, outcome -be
OR = (A/C) / (B/D) = AD/BC
Confidence interval - define.
Small/narrow CI
Large/broad CI
Indicator of your measurements precision.
Small/narrow CI - if same question asked again for different sample population then we are reasonably sure the results would be similar.
- 95% CI + 95% sure of similar result
Large/broad CI - less sure of result, ?increase sample size to increase confidence
CI is influenced by sample size
Relative Risk
1. Define
2. Used in?
3. How to calculate ?
- Risk of a certain event happening in one group vs another
- Used in cohort studies
Use 2x2 table to calculate again.
RR = a/(a+b) / c/(c+d)
Sensitivity - define
Specificity - define
Sensitivity - the ability of a test to detect disease
Specificity - the ability of a test to detect health
Calculation for sensitivity
Sensitivity = (TP/TP+FN) x 100
True positive
False negative
Tests with low sensitivity - waste of time / money
Calculation for specificity
Specificity = (TN / TN + FP) x 100
True negative
False positive
Positive predictive value
1. Define
2. Formula
The chance that if the test is positive, the patient has the disease.
PPV = (TP/TP+FP) x 100
When prevalence increases, so does the PPV.
Negative predictive value
1. Define
2. Formula
When tested negative for the disease they don’t have the disease
NPV = (TN / TN+FN) x 100
When prevalence increases, NPV decreases
Meta-analysis
Multiple, independent studies (preferably RCT’s) - same topic. Provides a more comprehensive and robust estimate of overall effect.
RCT
Subjects randomly allocated to design/control group (double blind) - to minimise bias.
GOLD STANDARD
Case-Controlled study
Retrospective.
Compares subjects with condition (cases) to those without (controls) to identifying factors which caused development of condition.
Cohort Study
Longitudinal research design.
Follows individuals over time to investigate relationships between certain exposures/characteristics and development of a specific outcome.
Forest plot - significance of 95% confidence interval including 1?
Not statistically significant findings
In a normally distributed population what % of the population lie:
1. +/- 1SD
2. +/- 2SD
3. +/- 3SD
- +/- 1SD 68%
- +/- 2SD 95%
- +/- 3SD 99.7%
Type 1 error
Reject a true null hypothesis
Type 2 error
Accept a false null hypothesis
