7 - Statistical Associations

Question 1

correlation

Answer 1

measures the association between two variables

- All we need is the covariance and standard deviations.

Question 2

Regression

Answer 2

• Allows to model the data
• Can be used with more than one explanatory variable
• Accounts for confounding
• Produces parameters and predictions.

Question 3

REGRESSION – canonical form

Answer 3

Y = a + bX + e

Question 4

REGRESSION: the error term

Answer 4

We expect a mean for e = 0 in a good model fitting.

Question 5

Causation in epidemiology

Answer 5

A cause is sufficient when it inevitably produces an outcome and is termed necessary if an outcome cannot develop in its absence.

Question 6

From association to causation

Answer 6

• study has an adequate sample size
• study is free of bias
• Adjustment for possible confounders has been done
• Possibility of reverse causation

Question 7

Causation Criteria (Gordis) 1

Answer 7

1. Temporal relationship
2. Strength of the association
3. Biologic plausibility
4. Dose–response relationship
5. Replication of the findings
6. Effect of removing the exposure
7. Extent to which alternate explanations have been considered
8. Specificity of the association
9. Consistency with other knowledge

Question 8

Temporal relationship

Answer 8

– The exposure precedes the outcome
• e.g. The smoking precedes the cancer.
• RCT cohort, cohort, case-control

Question 9

• Strength of the association

Answer 9

– Relative risk
• e.g. lung cancer and smoking RR = 10, for breast cancer and smoking RR = 1.5
– Stronger association is more likely to be causal, but a weak association can also be causal

Question 10

• Biologic plausibility

Answer 10

– The biological mechanism has to be plausible, e.g.
E.g in the statistical association between proximity to power lines and childhood leukaemia, there is no plausible mechanism

Question 11

• Dose–response relationship (biological gradient)

Answer 11

– If risk increases with increasing exposure, it supports the notion of a causal association

Question 12

• Replication of the findings

Answer 12

– Many studies show similar effect

Question 13

• Effect of removing the exposure (experiment)

Answer 13

Does the removal of the exposure alter the frequency of the outcome?

Question 14

• Extent to which alternate explanations have been considered (coherence)

Answer 14

The relationship found agrees with the current knowledge of the natural history/biology of the disease

Question 15

Specificity of the association

Answer 15

There is a one-to-one relationship between the exposure and outcome.
– This criterion is not applicable to all exposure-disease associations because a disease may be caused by several exposures, and an exposure may cause several diseases

Question 16

• Consistency with other knowledge

Answer 16

The same findings have been observed among different populations, using different study designs and at different times.