Part III Lecture 1

Question 1

What is a systematic error?

Answer 1

An error that affects all members of the study population

Question 2

What is a random error?

Answer 2

An error that varies from subject to subject

- mean = 0

Question 3

What is bias for continuous measurements?

Answer 3

The difference between the population mean of the measured exposure (ux) and the population mean of the true exposure (ut)
- but often the true exposure is unknown, so we take a gold standard measure

Question 4

Inter-method reliability

Answer 4

Comparing two different methods (e.g., gold standard vs less expensive)
either:
- where one is considered the true value
- where neither is the true value

Question 5

Intra-method reliability

Answer 5

Comparing repeated measurements by same instrument

Question 6

Validity coefficient vs reliability coefficient?

Answer 6

For intermethod reliability
p^2 or R^2 = variance of a/variance of b
- % of variance of exp1 explained by exp2

In the case of validity coefficient, a is the true parameter

In the case of reliability coefficient, a is one of the measure

Question 7

Measure of intramethod reliability?

Answer 7

Intra-class correlation
- % of the total variance due to between-person variation

Question 8

How do we measure the validity of categorical data?

Answer 8

Sensitivity, specificity

Question 9

How do we measure the reliability of categorical data?

Answer 9

Concordance (% agreement)
- but does not account for chance alone

Kappa statistic

• (%obs - %exp)/(1 - %exp)
• kappa declines as prevalence approaches 0 or 1

Question 10

What is the role of the coefficient of variation?

Answer 10

Express the reliability of repeated measurements of the same kind under the same conditions

• Standard deviation/mean x100%
• The smaller the value, the more precise

Question 11

What are the axes of a ROC curve?

Answer 11

x: false positive rate
y: true positive rate