# Lecture 6 - Chance, bias, confounding and true causal relationshiop Flashcards

What are measures of association?

- Identify causes of disease
- Estimate how much disease is due to an exposure
- Compare incidence of disease in two populations
- Common measures (relative risk, odds ratio, hazard ratio)

Relative Risk

ratio of the risk (incidence) of disease (the outcome) in those ‘exposed’ compared to the risk (incidence) of disease (the outcome) in those ‘not exposed’

RR = (incidence in exposed group)/(incidence in unexposed group)

What is the strongest study type in terms of evidence?

Meta-analysis of RCTs

What are two types of errors?

- Random Error

- Systematic Error (aka bias)

What is random error?

An error that decreases precision

What is systematic error?

An error that decreases validity

What are the three types of bias?

- Selection Bias
- Information Bias
- Confounding Factors

What is selection of bias?

Bias shown in selecting subjects

Usually results from comparative groups not coming from the same study base and not being representative of the populations they come from

What is information bias?

Bias shown in collecting information or measurements

What is the difference between chance and bias?

- Chance is caused by random error bias is caused by systematic error
- Errors from chance will dance each other out in the long run (w/ large sample size) bias won’t
- Chance leads to imprecise results while bias leads to inaccurate results

How do you asses the role of chance in a study?

- Hypothesis Testing

- Estimation

What is hypothesis testing?

- Use statistical test to examine the null hypothesis

- if p value is less than 0.05 then result is statistically significant

What is estimation?

- Uses statistical methods to estimate the range of values that is likely to include the true value
- It is associated with “confidence intervals” – if value corresponding to no effect falls outside interval then result is statistically significant

What are the benefits of Random Allocation (randomization)?

- Reduces bias in those selected for treatment (guarantees treatment assignment will not be based on patient’s prognosis)
- Prevents confounding (known and unknown potential confounders are evenly distributed)

What are the two conditions that must be met for confounding factors?

- Be associated with exposure (without being the consequence of exposure)
- Be associated with outcome (independently of exposure..not an intermediary)