Regression and bias

Question 1

Define chance? When occur?

Answer 1

Random error that produces different observations for replicate experiments or repeat samples
Occur before or after study

Question 2

State 2 ways prevent chance?

Answer 2

1) Ensuring a sufficiently large sample
2) Using confidence intervals and p-values

Question 3

What are confounding factors?

Answer 3

Factors distort results

Question 4

State 3 ways to prevent confounding at design stage?

Answer 4

Randomisation- fairly comparable
Restriction- eliminate variation
Matching- control group matches confounders in case group

Question 5

How prevent confounding at analysis stage?

Answer 5

Multivariable regression analysis
Stratified analysis- take weighted avg

Question 6

Define regression?

Answer 6

statistical method that shows the relationship between two or more variables

Develop model risk prediction clinical outcome
Estimate risk future incomes
State whether individual likely experience outcome or not
Prognostic and diagnostic

Question 7

Aim of regression?

Answer 7

Isolate single effect of single variable on clinical outcome

Question 8

Explain crude/adjusted effects?

Answer 8

Crude- doesn’t take account confounding variable
Adjusted- account confounding variable

Question 9

In simple linear regression what does y and x represent?

Answer 9

Explanatory variable (or the independent variable) always belongs on the x-axis
Response variable (or the dependent variable) always belongs on the y-axis

Positive negative or no

Question 10

In linear regression what quantifies association?

Answer 10

Slope coefficient (gradient)

Question 11

y=a+bc; where a = intercept, b = coefficient (slope) and x = predictor value

Answer 11

y=a+bc; where a = intercept, b = coefficient (slope) and x = predictor value

Question 12

What does multivariable regression do?

Answer 12

Adjust for all variable in model
y=a+b1X1 + b2X2 + b3X

Coefficients (b1, b2, b3…) show strength of association with each X value

Question 13

Define bias?

Answer 13

Bias is systematic error in design or conduct of a research study 🡪 result is different from the truth

Question 14

State 2 ways bias can be prevented?

Answer 14

Ensuring an appropriate selection of participants
Ensuring data is collected and measured correctly

Question 15

State 2 types bias?

Answer 15

Selection bias 🡪 problem with study population (selection study subject)
- factor affect recruitment or retention subjects
- e.g non response- cross sectional
differential loss- cohort study
selection cases- case control
Information bias 🡪 problem with information provided

Question 16

Is bias effected by population size?

Answer 16

No

Question 17

How can selection bias occur in:
Cross sectional
Cohort
Case control

Answer 17

Selection bias 🡪 problem with study population (selection study subject)
- factor affect recruitment or retention subjects
- e.g non response- cross sectional
differential loss- cohort study
selection cases- case control

Question 18

State 3 common source of information bias?

Answer 18

Researcher 🡪 observer bias
Participant 🡪 recall/respondent bias
Instruments 🡪 wrongly calibrated instruments