Bias

Question 1

What is a spurious relationship?

Answer 1

false cause-effect relationship between 2 variables when really there is a 3rd.

Question 2

What is the 3rd variable in the spurious relationship called?

Answer 2

confounder variables

Question 3

3 Criteria for being a confounder?

Answer 3

1. confounder associated with outcome
2. confounder associated with exposure
3. exposure DOES NOT cause outcome THROUGH confounder

Question 4

3 ways to reduce confounding when designing study

Answer 4

1. restriction
2. matching
3. randomization

Question 5

2 ways to reduce confounding when analyzing data

Answer 5

1. stratification

2. multivar analysis

Question 6

Restriction works how?

Answer 6

eliminate confounder by including it in BOTH groups.

Each group has smokers so smoking wont be confounder

Question 7

Matching works how?

Answer 7

Allow different levels of confounding variable, but match them between groups. Heavy smoker A, heavy smoker in B etc.

Question 8

Randomization works how?

Answer 8

Just allow people to join groups at random

Question 9

Randomization advantage?

Answer 9

Will probably be about the same in each group, and take care of confounders we didnt htink of.

Question 10

Randomization disadvantage?

Answer 10

Can get unlucky with distribution, and have more confounding on one side.

Can’t always randomly assign (can’t tell people to smoke, change their age etc).

Question 11

Stratification works how?

Answer 11

during analysis, break people into different groups based on confounding variable (age, smoking habits)

Question 12

Disadvantage to stratification?

Answer 12

have to defend a “cutoff” point you decide on for the strata

small sample size for each strata (might be very unequal)

Question 13

Multivar analysis works how?

Answer 13

statistical methods (multiple regression)

Question 14

Multivar analysis advantages?

Answer 14

inexpensive and can handle multiple confounders at once

Question 15

Selection bias - whats wrong with population?

Answer 15

selection doesnt mimick population cuz of bias

Question 16

Solution to recall bias?

Answer 16

choose groups equally likely to remember (both diseased0

mask subjects to study hypothesis

Find outside way besides asking to get info

Question 17

Social desirability bias is?

Answer 17

people over-report good

under-report bad about themeslves

Question 18

Way to get around social desirability?

Answer 18

randomized response (all yes’s, throw out half of the total question # of yes’s–if theres 100 total, throw out 50 yes’s]

need a BIG sample for this