Part III Lecture 3

Question 1

Ways error may occur?

Answer 1

• Faulty instrument
• Poor technique
• Poor recall
• Day to day variability
• Errors during data entry

Question 2

When is the exposure measurement error non-differential?

Answer 2

When there is equal bias and error variance (precision) in D+ and D- groups

Question 3

What explains the magnitude of the attenuation of the effect estimate when there is non-differential measurement error?

Answer 3

The correlation between the measured exposure and the true exposure

Question 4

What makes the impact of random measurement error differ?

Answer 4

Whether the error is in the outcome variable or the exposure variable

Question 5

Random measurement error in an exposure variable will… (continuous)

Answer 5

Bias the estimates of regression slope coefficient towards the null

Question 6

Random measurement error in an outcome variable will… (continuous)

Answer 6

Increase the standard error of the estimates and widen the confidence intervals (so less likely to be statistically significant)

Question 7

What is the impact of increasing sample size on measurement error?

Answer 7

Exposure variable: will only make the estimate more precisely wrong

Outcome variable: will minimize the impact of the measurement error

Question 8

What is the effect of non-differential misclassification of categorical outcome vx? How does it differ from continuous vx?

Answer 8

Bias estimate towards the null

in continuous: increases the standard error

Question 9

What is the classical error? Non-differential classical error results in?

Answer 9

• Measured values are distributed around the true population mean because of random variation
• Bias toward the null + loss in power

Question 10

What is Berkson’s error? Non-differential Berkson’s error results in?

Answer 10

• True exposures are distributed around the measured value
• When average exposure of a group is assigned as the exposure of that group
• No bias, but loss in power